Debate within USCV: USCV's VP Bruce O'Dell disagrees with USCV statement

[Time for change]

I have very mixed emotions about starting this new thread from a discussion that Bruce O'Dell posted yesterday in response to Ron Baiman's statement that was posted by eomer. The reason that I feel uncomfortable about this is that I fear that disagreement within USCV could be misinterpreted as providing some support for Mitofsky's "reluctant Bush responder" assertion, and consequently could lessen enthusiasm for trying to expose the 2004 election fraud, which I strongly believe did in fact occur, as indicated in this recent post of mine: http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=371726&mesg_id=371726. I believe that this whole issue needs a lot more public attention, not less, so I sincerely hope that this disagreement within USCV or my assistance in publicizing it does nothing to reduce our efforts to expose whatever happened in 2004.

And I also need to add that, as far as I can tell, Bruce O'Dell is not advocating for Mitofsky's position, but merely saying that USCV did not prove him wrong. I myself do not take a solid position on this controversy, as I do not understand the issues fully enough to do so. However, I do lean towards Ron Baiman's position, for reasons that I will make evident shortly after posting this thread, by re-posting a modification of a response I made yesterday on the other thread as soon as I saw Bruce O'Dell's statement.

That being said, despite my reservations I think that this deserves to be posted as a new thread because it involves a controversy that interested DUers should be familiar with, and it probably would receive very little attention being hidden away on the other thread. So here is Bruce O'Dell's statement, word for word as it appeared yesterday:


I'm Bruce O'Dell - the Vice President and co-founder of US Count Votes.

With all due respect, I believe Ron's interpretation of Mitofsky's findings is fundamentally mistaken, and so is the USCV Working Paper, first published May 12.

After unsuccessfully working within US Count Votes to revise or retract the Working Paper that a minority of the USCV membership recently published, I see no alternative but to publicly challenge the report’s methodology and conclusions.

The key argument of the USCV Working Paper is that Edison/Mitofsky’s exit poll data cannot be explained without either (1) highly improbable patterns of exit poll participation between Kerry and Bush supporters that vary significantly depending on the partisanship of the precinct in a way that is impossible to explain, or (2) vote fraud. Since they rule out the first explanation, the authors of the Working Paper believe they have made the case that widespread vote fraud must have actually occurred.

However, a closer look at the data they cite in their report reveals that Kerry and Bush supporter exit poll response rates actually did not vary significantly by precinct partisanship. Systematic exit poll bias cannot be ruled out as an explanation of the 2004 Presidential exit poll discrepancy – nor can widespread vote count corruption. The case for fraud is still unproven, and I believe will never be able to be proven through exit poll analysis alone.

The fact that I chose not to endorse the USCV Working Paper should be a clear indication that I do not support its central thesis, and in fact believe that the simulation data they cite refutes the Working Paper’s conclusions.

I am not a statistician, but as a computer systems architect, I create mathematical models to simulate the performance of large-scale computer systems, and mathematical simulation of the cost and efficiency of business processes is a significant part of my consulting practice. My own election simulation results are cited on pp. 9 -10 and in Appendix G of the May 12th Working Paper; as the creator of the only USCV simulation which accurately reproduces aggregate Mean WPE, Median WPE and participation rate data from the E/M January report, I feel an obligation to ensure that my work is correctly interpreted.

I can show that several of the USCV election simulation programs are flawed, and that when the Liddle Bias Index is applied to the “USCV O’Dell simulation” data cited in the Working Paper, it produces results consistent with those recently reported by Warren Mitofsky for the E/M data as a whole.

I respect Ron's opinion, but his insistence on using aggregate WPE as a tool to interpret poll response bias (or vote fraud) is mistaken. His analysis of the Liddle Bias Index is also off-target. Liddle's Bias Index is an inherently superior metric to WPE, and analyses based on aggregate WPE are highly misleading.

I've written a paper that addressed this issues in detail, that can be found at www.digitalagility.com/data/ODell_Response_to_USCV_Work... .

If anyone can show me where I'm wrong, I'll be the first to admit it.

I'm disappointed that I was not able to resolve our disagreement within USCV, but I simply cannot allow a fundamental misinterpretation of my data - the USCV O'Dell simulator they cite in their paper - to continue to go unchallenged.

In addition responding to this posting, please feel free to contact me at my email address at USCV, [e-mail] - or at my corporate email address at [e-mail] if you have any questions.

[firefox]

1. It's like 9/11

Nobody was fired. There are no criminal prosecutions. There is no investigation of a serious nature by government. The media does not call for an investigation or raise key questions and related facts.

The government wants it buried while the rape of the treasury and the destruction of the general welfare continue. The real powers that be removed the paper trail for a purpose. Heads should roll just for that. It is absurd on the face of it not to have a record of votes cast and both parties were in on the destruction of having such records as a basic requirement.

Even if the election were not stolen, although I strongly believe it was, there should be outrage from the public. There is plenty of stink now without stirring the shit with a stick/probe.

Impeach Bu$h and replace Congress.

[autorank]

2. Request: A list of non USCV Statisticians & No more dirty laundry here

These internal politics and much too public disputes, i.e., the O'Dell memo, are truly not helpful and display a degree of non professionalism and internal rancor that are simply irrelevant to the task at hand.

This type of dispute is best handled internally, with private correspondence.

DU is being used as a playing field to discredit some and prop up others.

Quite frankly, it's juvenile, particularly this post by O'Dell.

To wit, I'd like some links to statisticians or organizations who are doing statistics or the individual efforts of USCV members relating to election fraud. I'm quite happy with what I've got here at DU with all our regular election fraud workers and TIA's superb and productive analysis. I'd just like to see more rather than focus on arcane debates that seem focused only on salvaging the final National Exit Poll.

I am quite familiar with DU and I've never seen a forum where an organization comes to discredit other members of the organization through "new users" coming just for that purpose, it would seem; and I've never seen a forum where people show up to engage in nasty, personal attacks on a key forum users.

It's rude and unnecessary. Stop it and do something about election fraud, an ongoing request I've made to the visitors, and a sincere one.

[Internut]

5. From his post, I understand that he did

try to handle it with private correspondence. But when Ron Baiman et al insisted on going public with the paper that incorporated Bruce O'Dell's work without taking into account his objections, that left Bruce only one recourse - to go public with the objections.

I don't understand why you think that Bruce O'Dell should have shut up when his colleagues published a paper that Bruce considers (I quote) "fundamentally mistaken" and "highly misleading" and used his work inside it. There is such a thing as academic honesty, which requires one to publish results even if they do not fit one's agenda.

[autorank]

7. This is not the place to do it. It's about elections not internal

politics. Re-read my message, it's clear and accurate that DU is being used here for internal politics. No other forum has to put up with this.

[kster]

389. THIS GUY REALLY KNOWS HIS NUMBERS

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=104x3858119

[mgr]

6. That is one of the problems with transparency n/t

[autorank]

9. This is a high level version of the Jerry Springer show.

Transparency does not inure to internal politics of an organization, sniping if you will, being shared on our forum which is more action oriented. This is really tedious. What's next, DNC posting internal political battles in the Democratic Party forum? Or maybe users posting every single thought, reservation, fantasy as they read our posts.

No, this is not the place for this. It's simply bad form.

[mgr]

18. No one said democracy was neat

I understand your feelings, but much of this was anticipated in Febble's earlier posts. Being from academia (and married to a Friend), I can tell you that this is nothing, but characteristic of most academic allegiances where having it right comes second to consensus; and in this instance is appropriate for this forum. Open debate fosters greater understanding of the issues, and makes for a better nuanced understanding of what is at stake. You can pick your heroes, see where your position actually stands, and get an idea of what the actual issues are. I don't see any mudslinging or ad hominen attacks--it's quite the model for DU.


Mike

[autorank]

20. Here is the best line I've heard about academic politics:

"The politics are so vicious because the stakes are so small."
(anonymous, by request)

This helped me understand previous dealings with the postsecondary world.

In this case,Mike, the stakes are very large. This is appropriate to the USCV board, another board, but not DU in my opinion, and that's what I'm expressing.

Paralysis by analysis is the MO lately around here and it doesn't move things further down the road. As landshark pointed out, there is more than a prima facia case, considering all the evidence, for a fraud investigation. That's what we should be pushing for.

I'm a great admirer of the Society of Friends and you're a lucky guy. From my recollection, and it's been some time, when a group decision is made, they move forward with impressive resolve and unity.

The decision here is that fraud took place. We can go on endlessly around the margins but that only makes us look disunited and somewhat moribund, despite our small number.

[Febble]

24. OK, I get it.

In which case the title of this forum needs to be changed. It is not "2004 Election Results and Discussion"

It is "Forum for those who believe the election was stolen to assemble any evidence for the prosecution, regardless of its persuasiveness".

It would have saved some of us a lot of time and heartache.

And I would wish you the best of luck.

[mgr]

29. Much of what's being done is outside of this forum n/t

[Melissa G]

46. OK Mike, This is where spouses of Quakers often miss the point

I know because my husband makes this same error.
A major problem in Quaker process is the framing of the question being considered. Quaker process can be arduous and too focused for broad questions that need a lot of precise answers.
As I am sure you know, there is a big difference between Consensus and Unity. Consensus is what Friends strive for and many Friends have been known to "Stand Aside" to let the group move forward in Unity. Consensus does not necessarily mean that everyone agrees, just that no one's conscience must keep them from blocking group action or the "sense of the meeting".
How does that apply to this discussion? There is a strong consensus that the election was stolen by ohhh so many ways. This is overwhelmingly obvious to the group that is convened here. It is the focus of our "Concern". Folks understand limited time and energy and want to make a difference in this concern. They reasonably object to framing that they believe gets the discussion off topic and into eddies of irrelevancies. While this explanation is a religious comparison, this consensus about fraud is based on facts not beliefs.
Our concerns about wasting energy on academic points instead of focusing our limited life energies working for justice are valid.
God only gives us so much breath.

Febble's paper and Odell's simulator while academically interesting are academic eddies that slow process. I also believe that they allow for political misuse of their academic arguments.
edit typooo

[autorank]

48. Bless MelissaG: "allow for political misuse of their academic argument"

What many have struggled to say in too many words, you summarize in a phrase!

[Time for change]

52. Good point Melissa, but I have another point

I believe that everyone posting here, including those who disagree with USCV's recent paper, agree that election reform is of great importance, and we should be working towards that. Academic disagreements such as this should not interfere with that, because election reform advocates who are not interested in these discussions don't need to spend any time reading about them.

However, I believe that it is also very important for some of us to work towards exposing what happened in 2004. I believe this because at this time the American public, the MSM, and Congress are way too apathetic about election reform. As long as that remains the case, our election reform efforts are not very likely to be successful. I believe that honest academic discussions such as this are useful in moving us towards that goal.

[Melissa G]

79. I agree with everything you said TFC

my only concern is when academic discussions are used to cloud an issue of greater import. Like a stolen election..

[Time for change]

81. I agree with you on that

Whenever I see something that I interpret to be an effort to cloud this issue I get very upset, and sometimes I've even said some things that I probably shouldn't have said in response to that kind of stuff.

But I'm almost positive that that is not what is going on here. See my post # 76, below, for details on that.

[mgr]

60. Absolutely correct

I was raised Congregationalist, so going over to Friend was not hard. I've stood aside on several occasions. But what needs understanding is that neither unity or consensus on the matter may have been reached in regards to USCV's releasing the paper. I am not in consensus that exit polling irregularities are necessary to address fraud in the election, and are quite likely harmful to our position within the general public, and because of that I don't view Febble's correction term, and O'Dell's simulator as minor academic eddies, but harbingers of how the argument will be addressed within the wider public.

(By the way, why is it the husbands but not the wives? <smile> )delete

Mike

[Febble]

62. I was a Quaker

when I married a catholic man.

I found the "feeling of the Meeting" stuff frustrating at times - good for theological issues, terrible for fundraising drives.

And I liked the rigor and the dialectic of a church that knew how to argue - even if it comes to the wrong answer a lot of the time (but I am good at finding logical loopholes). So I converted.

But where I absolutely agree with you is that poor arguments are actually damaging to a case. That's why I keep bringing up the proportionally spaced font. The case for fair, secure, auditable elections is completely irrefutable. The exit poll evidence is anything but.

Perhaps most important of all - arguments are strongest when they don't represent a vested interest. Fair elections matter to all Americans, not just the ones didn't vote for Bush, and especially not just to those that think he didn't win.

And what about voter suppression? Expecially of African American voters. It was massive, it was systematic, and would have had virtually no effect on the exit polls (it may have marginally affected weightings).

If I was a conspiracy theorist, I would suggest that the exit polls were rigged to look like Kerry won, so that people would concentrate on fraud hypotheses and forget about the scandal of voting machine rationing to predominantly African American precincts in Columbus, or the provisional votes scandal in predominantly African American precincts of Cuyahoga county. Kerry lost tens of thousands of votes in Ohio not because of fraud (although that may have happened too) but because a partisan SoS did his utmost to ensure that people likely to vote for Kerry had less opportunity to get their votes cast at all, let alone counted.

Compared to that evidence, a rough survey of voters leaving 1% of the nation's precincts is at best a distraction, and at worst a trap. IMO.

[mgr]

63. Voter suppression is the partisan issue

I made the argument that long lines in Florida may reflect an attempt to suppress the jewish vote--as it occurred predominantly in the southeast coastal counties--Volusia, Palm Beach, Dade.

Mike

[Febble]

69. Correction.

I had been under the impression that voting machine allocation in Franklin County Ohio was under the direction of the Republican director of the Board of Elections, Matthew Damschroder. I have now been informed that allocation was done by the Democratic chairman, William Anthony.

It was a mind-boggling piece of incompetence in my view, and cost Kerry votes. But once again, an apparent conspiracy turns out to be a cock-up.

Thanks, Mike.

[LightningFlash]

152. Actually, Febble, you seem to have missed the point alltogether.

http://freepress.org/departments/display/19/2005/1284

"In a letter dated October 5, 2004 to Republican Chair of the Cuyahoga County Board of Elections Robert Bennett, Blackwell specifically threatened removal of any board member who refused to follow his direct orders. The threat undermines Republican arguments that the election was fair because both Democrats and Republicans serve on election boards. "Be advised that your actions are not in compliance with Ohio law and further failure to comply with my lawful directives will result in official action, which may include removal of the Board and its Director," Blackwell wrote Bennett.

Under Ohio law, all election board members serve at Secretary of State Blackwell's pleasure. Cuyahoga Bureau of Elections director Michael Vu mentioned the letter at a Congressional hearing staged at the Ohio statehouse by Republican Congressman Bob Ney. Ney brought the hearing to Columbus in part because Blackwell refused to testify in Washington. The hearing was highlighted by angry, bitter exchanges between Blackwell and US Rep. Stephanie Tubbs Jones, who co-introduced (with Senator Barbara Boxer of California) the historic Congressional resolution challenging the seating of the Ohio Electoral College delegation for Bush."

Bi-partisanship was actually ruled out conclusively, Febble. As B.O.E. members were instructed personally to follow Blackwell's direct orders for the whole period, including when dealing with machines, or face punishment.

[Febble]

157. I was talking about Franklin County

A poster above corrected me. I had understood that the Republican director of the BoE in Franklin county was responsible for the formula used for machine allocation.

It appears that it was a Democrat, the chair of the BoE.

Whoever it was, I stand by my finding that shortage of voting machines in Columbus disenfranchised more Kerry voters than Bush voters, and that at best this was due to quite incredibly poor management of what should not even be a scarce resource.

Check my paper if you want more details. It supports your case.

http://uscountvotes.org/index.php?option=com_content&task=view&id=65&Itemid=63

[Melissa G]

77. Hi Febble, I was raised catholic and later became a Friend by

convincement as an adult. I'm Hispanic and married to a Jew, as well as the daughter of a civil rights activist attorney. I know about arguing and I know about the frustrations and limitations of the sense of the meeting discussions trying to do mundane things like buy and sell real estate and building buildings.. wow do I know frustration...
You wrote...
Perhaps most important of all - arguments are strongest when they don't represent a vested interest. Fair elections matter to all Americans, not just the ones didn't vote for Bush, and especially not just to those that think he didn't win.

Agreed.
However, you represent a vested interest.. ( yourself, and your employer Mitofsky) You wrote a paper which you are getting a lot of attention for. I have seen you write that your theory has little impact on whether or not fraud occurred, yet I believe you are being used by others such as Mitofsky to deflect from the importance of releasing the data so others, lest vested, can analyze where the problems are. I am interested in who won. I have yet to see the Bush padding have a reasonable explanation.
Best,
Melissa

[OnTheOtherHand]

83. sigh

Melissa,

I'm a little bit frustrated by this, because I have seen Febble go through it quite a few times. I am not employed by Mitofsky, and I generally agree with Febble on her major points. I thought she was honest before she picked up contract work for Mitofsky, and I still think she is. In my experience she is more ruthless with her own work than most of us manage to be. Also, I don't think Mitofsky can release all the data. My impression is that most survey researchers don't think he should release all the data, most other people haven't given it a moment's thought, so it's not obvious why he would need cover from a British psychology grad student (albeit a very, very smart one).

All that said, I agree that the more information released, the better, as long as legitimate privacy concerns are respected. I don't mean to quarrel or quibble over exactly how vested Febble is -- it really shouldn't matter that much. We can read her arguments and form our own opinions. It just seems to me that the attention given to Febble's vested interests is disproportionate. Frankly, anyone who has ever expressed a strong opinion about this issue probably kinda wants that opinion to be proven right.

[Melissa G]

84. OTOH, sigh all you like.. You are reading my post differently

than intended. I am not saying Febble is not honest. I am not impugning her motives. I am not doing crit on the technicalities of her work (not qualified even if I wanted to...which i don't)
You write..
Also, I don't think Mitofsky can release all the data. My impression is that most survey researchers don't think he should release all the data, most other people haven't given it a moment's thought,
mg..
I'm not suggesting he release all the data...as you well know there are ways for him to release the data that would cover the objections of most..

More from you..
so it's not obvious why he would need cover from a British psychology grad student (albeit a very, very smart one).

mg again..
If you really think this then you are not very savvy about the concepts of political cover and red herrings..
I agree that Febble is very very smart about some things. I have concerns about others but I assume no malicious intent by her. I am only worried about certain unintended consequences.
edit for clarity

[OnTheOtherHand]

93. honestly, I think you could do better

Ordinarily, I would agree and accept that you haven't questioned Febble's honesty or impugned her motives. But she has gotten pretty well ripped in this forum, and your post continued the pattern of focusing more attention on her "vested interest" than anyone else's (well, occasionally people go after mine).

EDIT: Sorry, that was unclear. I am willing to agree and accept that you haven't intended to impugn her motives. I just don't think that's good enough.

Can we clarify the data issue? My impression is that a lot of people on the forum still are calling for Mitofsky to release all the data, but I don't know what your position is on this. (You may well have stated it -- there are a lot of posts!) What do you think would constitute adequate release? You may not be entirely sure -- I'm not -- but it would help if we could drift toward agreement on this.

There's not much point in debating how savvy I am. But for what it's worth, if most citizens or most public opinion researchers or most people outside DU thought Mitofsky should release something he hasn't, then I don't see how most citizens or public opinion researchers or people outside DU would be swayed by Febble. It's true that her arguments could be used in ways she doesn't intend. But if we limit ourselves to arguments that couldn't be misused, then we can't actually say very much.

[Melissa G]

129. OTOH, What is it you think i could do better about?

On the data..people I trust who are more knowledgeable than i have told me there are ways to release the data that would cover the confidentiality and other concerns..I tend to believe them. Do you not think this is the case? and if not why?

My statement re your savvy was not about data release but by the use of Febble by Mitofsky...
I do a fair bit of civic activism and local politics and this is a not uncommon ploy..

Mitofsky has a problem..Short version is people are attacking him about the exit poll data, validity of his work,importance of him releasing the data, yada yada yada..

He needs a distraction...
Enter Bright Febble known Dem poster with a Very Clever Formula. Febble is a good soul and has been the first to admit that her formula does not say whether or not there was fraud. She goes on a bit about how misunderstood she is and how proud she is of her formula but if you read around that she readily admits its limitations.

Mitofsky befriends Febble and Buddy Mystery Pollster is also involved in the Febble friendship (clear disclosure here..I am not at all trusting or of either Mitofsky or MP's motives). Not only does he flatter the heck out of her. He goes as far as hiring her and contractually obligating her to him. This makes her less likely to object to what her work is being used for.

He then involves her in a totally BOGUS imaginary peer review process which he presents with much fanfare at the AAPOR convention in Miami. He uses Febble's pretty picture to obscure the Fact that there is NO Actual Explanation for the Difference between E/M's data and the Recorded Vote.. Nice Picture but NO Accounting.

He then gets ESI to put out this silly press release...
http://biz.yahoo.com/prnews/050514/nysa019.html?.v=3
that he gets to wave about and send out on the wires all over the world... Nothing to see here folks. Move Along..

Instant Political Cover.
This is what I object to.

Edit for spacing

[Melissa G]

132. First part of annoying , misleading, ESI press release

Press Release Source: Election Science Institute


Ohio Exit Polls 'Not a Smoking Gun' for Fraud, Study Says
Saturday May 14, 6:54 pm ET


MIAMI, May 14 /PRNewswire/ -- A just released analysis confirms pollster Warren Mitofsky's assertion that the exit polls that put John Kerry ahead of George Bush in Ohio on Election Day 2004 do not necessarily indicate that there was fraud in the Ohio election.
http://biz.yahoo.com/prnews/050514/nysa019.html?.v=3

[Time for change]

137. That statement in the press release by Dr. Scheuren

"voting patterns were consistent with past results and consistent with exit poll results across precincts"

That's absurd. How can he say that voting patterns were consistent with exit poll results when there was a huge difference between them?

[Melissa G]

138. Exactly TFC .. Silly , Misleading and Annoying BS... n/t

[OnTheOtherHand]

160. I'm not sure, but I suspect that meant

that the extent of error didn't vary systematically with partisanship -- so, "consistent... across precincts" in the sense that the error was pretty even.

I will need to read some actual work from ESI (so far only the PowerPoint has been posted, apparently) before I form any opinion on that. I do know that Fritz Scheuren is smart and experienced, so I look forward to looking at the work.

[Time for change]

167. Whatever he meant, the statement he makes implies to most listeners

Oh, we looked at the discrepancy between the exit polls and the official vote count scientifically, and what we found was that everything is explained now. The so-called discrepancy doesn't need to be explained any more because everything is "consistent".

That's my translation, and I'll bet you anything that that's the way that the great majority of people who read it took it.

In my view, MalissaG's point is right on, in that these kinds of things contribute greatly to a misunderstanding of the situation by the public and add to the public complacency about our election system. The statement was grossly misleading.

[OnTheOtherHand]

169. I'm in no position to agree that the statement was misleading

Bottom line, I think what most people will take away is that Scheuren went looking for exit poll evidence of fraud in Ohio and didn't find any, which seems to be true (i.e., that he didn't find any). The man is president of the American Statistical Association, and so I hesitate to assume that he is mischaracterizing his own results. "Not a smoking gun" seems pretty far from an "all clear" to me; I'm not sure what they could have said that wouldn't be misinterpreted by someone.

As for the rest, it seems pertinent to me that the press release goes on to quote Steven Hertzberg: "The fact that there is debate over this at all shows that we need elections to be more transparent, more accountable, more auditable." I would have thought it was a good idea for election reform advocates to emphasize that point.

[Time for change]

181. Yes, the comment about the need for election reform is good

I was referring to Dr. Scheuren's comment in the press release, not any of his findings, which I am not familiar with. I think that "voting patterns were consistent with ... exit poll results across precincts" implies much more than that he didn't find evidence of fraud.

[Melissa G]

198. The comment about the need for election reform is what

folks put at the end of attack articles to try to appear 'fair and balanced'. The bulk of this Press Release is about the red herring of rBr.
The PR says a statistical heavyweight is backing Mitofsky with "analysis (which) confirms pollster Warren Mitofsky's assertion that the exit polls that put John Kerry ahead of George Bush in Ohio on Election Day 2004 do not necessarily indicate that there was fraud in the Ohio election.

CYA Mitofsky...

[Febble]

144. This is complete surmise on your part, Melissa

If you do civic activism you will know the dismay when something you know happened one way is reported in the press as happening in quite another.

Well, that's what's happened here.

And for the record:

Neither flattery nor abuse will change what I think. Evidence and logic might.

And no peer-review process, bogus or otherwise, was involved in what Mitofsky presented at AAPOR.

[Melissa G]

196. Yes, I know that experience but then I strongly object to the

misrepresentation and get a retraction or another story on the issue rather than embrace it as Mitofsky has done. Also this was not what was reported but an ESI PRESS RELEASE that Mitofsky could have objected to and possibly influenced.
Mitofsky released data to you, ESI, and Mark Blumenthal that others do not have access to. He then gets to use this ESI press release to say that all the unanswered questions were accounted for in Miami. As I said BOGUS. I intend you no abuse nor have I given any. Please see my pm and do not conflate me with others.

Edit for clarity and emphasis

[OnTheOtherHand]

163. some cynicism is actually naive

First of all: On the data front, I think blurring could work, as proposed by... uh oh, Fritz Scheuren at ESI. (Or I should say "might," since I still haven't gotten to "vet" what ESI actually did.) I'm open to discussing other things. Do you know what methods your people have in mind?

Now: the idea that Mitofsky scored huge political points at AAPOR by (correctly) crediting some of his work to Febble would probably seem silly to anyone who has belonged to that sort of organization. (Rightly or wrongly, the idea that Mystery Pollster, who works for Democratic campaigns, has ulterior motives in questioning the evidence for fraud also seems silly to many of us.)

If I understand rightly, you have no intention of engaging the statistical substance of ESI's work at any point, because you don't do statistical analysis. Therefore, I have a hard time fathoming your confidence in critiquing work that you will never attempt to understand. It's not as if these folks are creation scientists.

[Melissa G]

186. Feel free to put me on ignore.. I still will likely read you even

though I find your condescending tone to both myself and Febble annoying. Being Quaker, and in living in a very cool major university town that many folks want to live in after graduation means that my list of friends and acquaintances is chock full of Ph D's, doctors, lawyers, software geeks and other academics . Just because I cannot fully follow the intricacies of all the math does not mean I cannot follow the stream of concepts.
Ph D's who are are insulting, generally stay on the far reaches of my acquaintance so, as I said, feel to put me on your ignore list. If you think folks who don't do intricate math are stupid, well, enjoy your limited world view.

[Melissa G]

188. I believe blurring is what folks are talking about but I think some

make different distinctions than others. As long as it works and we can get data out for folks to analyze it is beyond my need to get technical about. We all have our work in the world ..this technicality is not mine.

[Melissa G]

197. OTOH, you assume a different audience than I do

IMHO, politically AAPOR is merely the backdrop for the ESI press release. The tag at the end from Steven is what often gets put on articles that try to appear 'fair and balanced'. The bulk of the release implies that a statistical heavyweight is backing Mitofsky to say that
'analysis confirms pollster Warren Mitofsky's assertion that the exit polls that put John Kerry ahead of George Bush in Ohio on Election Day 2004 do not necessarily indicate that there was fraud in the Ohio election."
The Audience is those who read newspapers or watch TV and will only get the McNugget version.

[OnTheOtherHand]

211. (trying a two- or three-in-one response)

It's in no way condescending to Febble to say that Mitofsky doesn't need "cover" from her at AAPOR. In my opinion what he needed from Febble was help with a statistical problem that she understood better than he did.

It's not PR that a statistical heavyweight found no evidence of fraud in the Ohio exit polls, it's the simple truth as far as I can tell. If you talk with Fritz Scheuren and are convinced that he was put up by Mitofsky to do CYA, I will be interested to learn why you believe so. If you talk with Steven and are convinced that he is a Rovian running dog, or whatever you think he is, I will be interested to learn why you think that.

Otherwise, I wouldn't have thought that the Quaker tradition would impel you to challenge factual claims on political grounds. I am truly puzzled about your motivations.

[Melissa G]

214. Talk to Febble.. She'll 'splain it to you..

I have no trouble communicating with her but i find you pretty consistently seem to misunderstand my communications. I don't think you get much of a bigger picture than the occasional tree you insist on staring at.
This is a pretty uninteresting conversation but for the record...
Paragraph one talk to Febble.

paragraph two.. go try TIA's numerous links to get you out of your 'simple truth'. To me, it is a fact that ESI"s press release was misleading and that Mitofsky used AAPOR to say that his presentation explained discrepancies which were not explained. You and I may disagree about this. I am unlikely to talk to Fritz Scheuren unless he shows up here. I have blogged with Steven as he is a member here and I am frankly disappointed with what he is doing. If he shows up again I suspect several folks would express their disagreement.

Paragraph three. My motivations are pretty transparent. I see a Mac Truck worth of information that says our democracy got heisted. I am amazed that you are so lost in the maze of rBr that you don't see it.
Go visit TIA over at
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x375366
and maybe he can lead you out.

If you want more links about why I believe we have major election FraWd, pm me, I'll be happy to share them.
Holding you In the Light,
Melissa

[Febble]

87. Melissa, please read my post below

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=374482&mesg_id=374867&page=

Mitofsky is not "my employer" in any general sense.

As a result of the totally independent work that I did, and which has been commented on here, Mitofsky reanalysed his own data. For that matter, a number of the original USCV authors my work seriously enough to reconsider their conclusions.

Mitofsky decided to present the results of his new analyes at AAPOR. A week before his presentation, he engaged me to run some more.

That does indeed give me a small vested interest - it is difficult to be sure you can remain entirely objective when your work is done for an employer, rather than for yourself. However, it is a very minimal contract, and frankly, my professional reputation as a honest scientist is worth far more to me, even in sheer monetary terms, than this bit of piecework.

However, the work of mine referred to here, and by Bruce O'Dell in particular, was done entirely independently.

And if have evidence that I am being used by Mitofsky to "deflect from the importance of releasing the data" I'd like you to direct me to it.

[Melissa G]

131. Hi Febble, I understand everything you just wrote.

Sorry, I had not read it when I posted #93 above or I would have responded directly to you rather than OTOH.
If my answer to him does not answer your question, I would be happy to clarify.
Best,
Melissa

edit commas

[Melissa G]

136. Sorry Febble, it was post 129. too tired... n/t

[Awsi Dooger]

427. Thank you Feeble; that has always been my contention

We are frothing to obsess over non-existent high tech fraud while our opponents merely spread and perfect the same tired vote suppressing tactics that have prevailed for decades. I wrote that many times on DU before the election.

I backed out of this fraud debate in early December for reasons you stated in an earlier post in this thread; the DU forums regarding November 2004 had become conviction only and any defense argument was dealt with via extreme ridicule. I have an Excel spreadsheet evaluating more than 200 carefully isolated bellwether counties nationwide going back more than a decade, every type of voting apparatus and representing all 50 states. When I slowly inputted the 2004 data as it became available for those counties it was landslide obvious there had been a small but critical shift toward Bush and the GOP. It got to the point I could basically predict the number within hundredths simply by looking at numbers from 2000 and 2002.

I want to thank you for the incredible contribution. I read about your discoveries via Mystery Pollster two weeks ago and have finally looked at your work. Fantastic! Dr. Fritz Scheuren also deserves praise for comparing this year's numbers to those of previous elections, instead of blindly accepting early exit poll numbers which make zilch sense in relation to the long established and highly predictable partisanship of a state or precinct. After looking at studies from Dr. Steven Freeman and others, I came to the conclusion I could tell them early exit polls from Utah had Kerry leading by 6 points and they would do nothing but punch that +6 in a model and compare to the actual vote.

[Melissa G]

75. Hi Mike , are you on the USCV stats list and so privy to their process?

I'm sure they likely don't proceed after the manner of Friends..
This forum also does not have that process in place. As you know, Friends process of discernment is to try to reach clarity on the leading of Spirit. Folks only generally speak if they have a new piece of information that adds to the search for Truth.
(Are you sure you read my post? I used gender neutral "spouses", only using husband in reference to myself..maybe if you read posts more carefully you would see how others believe the election was stolen.
The Stolen Election looks as least as big as the States of Florida and Ohio to me? maybe bigger..)

[mgr]

232. I'm using the terms loosely

Sorry Melissa, I'm a little late in replying, I only have internet access at work.

I once was privy to how academia works. There are often issues that require either consensus or unity. The thing to keep in front of us when discussing USCV, is that all its members, think that some sort of fraudulent actions occurred with the election, so all probably are at consensus on that. The minority paper and attribution issue only suggests that they are not in consensus on how to proceed in addressing the matter. As an observer, I don't see the matter as that they are at loggerheads--with the appropriate apologies, mea culpas, and agreements to respect others, they are capable as working as one. I've seen the same with Friends, and on some of the civil disobedence action groups I used to be with, although far more subdued and polite than what transpired here.

Lately, I have been reading posts fast rather than quick, and caught the spouse reference in the header on the second read through. I had caught only the text when responding (and indicated my mistake in the interests of honesty, so someone can see what my error was). Please accept the apology.

Mike

[Time for change]

8. I'm sorry you feel that way about it autorank

I view this differently than you. I firmly believe that both parties to this dispute (Ron Baiman and Bruce O'Dell) are strongly committed to improving our election system -- that's why they've devoted so much time to working for USCV.

Also, I don't at all see the purpose of their argument being to discredit each other. They have different opinions about this issue, and they are debating it with each other, just as many other issues are debated on DU. Notwithstanding my preface to this thread, I feel that it is good for us DUers to be exposed to differing points of view, because I think that we can learn from then. Although I too strongly believe that John Kerry really won the 2004 election, there are many issues relating to this that have not been settled, and I think that we can learn from discussions such as this and use them to make progress towards exposing what really happened in the 2004 election.

[autorank]

10. Paralysis by analysis...perhaps by design.

USCV has it's own web site. Why not do it there? or start a Yahoo group? Respectfully, I've been around here at DU longer than you have and I simply don't see this type of stuff go on. You can participate if you want but what I'd really be interested in from you or anyone else, is the list of statisticians I asked for with any links. Thanks in advance.

[Time for change]

11. I don't have a list of non-USCV statisticians

I'd give it to you if I did. But I don't know who in the DU are statisticians. I could guess, but your guess is as good as mine.

[Febble]

3. I really appreciate your posting this

I pasted it into the Wisconsin thread, but didn't want to post it as a headline topic myself, for perhaps obvious reasons.

But I do recommend people read Bruce's paper. It is very readable, and stirring stuff. He is certainly not advocating complacency over the election - quite the reverse. It ends with a potent call to arms.

But if anyone wants to understand what might have gone on in the 2004 election, and what the exit polls might or might not be able to tell us about it, it is really worth reading, whether you find yourself in agreement with his conclusions or not. And in fairness to Bruce, if people are going to read the Baiman Dopp paper, which is based on his simulator, they should really read what the creator of the simulator has to say as well.

[Time for change]

4. Response to Bruce O'Dell criticism of USCV response to Mitofsky AAPOR talk

Thank you very much for sharing this information with us. I am greatly interested in this issue, and I would very much like to understand it better. I do agree with you that the Liddle "bias index" is a superior measurement of sampling error, compared to WPE. Yet, it is not totally clear to me from reading your post what your position is on certain things, so if you have the time I would very much appreciate your view of the following points:

1) Here are some ideas that I recently posted explaining why I believe that Mitofski's "non-significant" slope not only doesn't support rBr but actually argues against it:
http://www.democraticunderground.com/discuss/duboard.ph...
I would be interested in your response to my first two points.

2) Later on in the same thread, in a discussion with Febble, I make the broader point of why I don't believe that a significantly positive slope is a good argument against fraud:
http://www.democraticunderground.com/discuss/duboard.ph...

3) It is not clear to me what you believe to be the fatal flaws in USCV's most recent publication. My understanding is that you feel that at least part of the problem is that it is based on some incorrect assumptions.

But here's what I don't get about that: It seems to me that assumptions or even simulations aren't even necessary. Considering the mean WPE in the Bush strongholds (>=80%) of -10.0, and the median of -5.8, that means that the 20 precincts with the most negative WPE in the Bush stronghold would have at the very least (least negative that is) a mean WPE of -14.2. Making the conservative estimate of a mean WPE of -14.2 and a Bush vote count of 85%, and assuming a total sampling rate of 56%, one could show through some algebraic calculations that the Kerry sampling rate would have to be 83% (for this group of 20 precincts as a whole), the Bush rate would have to be 51%, and alpha would be 1.61. No assumptions necessary, except that the data Mitofsky provided are correct. Do you not think that that is an implausible scenario in the absence of election fraud in some of those precincts?

4) I've looked very hard at Mitofski's plot of ln(alpha) vs. Bush vote count, and it doesn't look non-significant to me. Precincts with ln(alpha)>.5 and Bush vote count>=80% look very numerous to me, whereas the number of precincts with ln(alpha)>.5 and Bush vote count<=20% looks very sparse. Are we sure this line is not significantly positive? And furthermore, it certainly seems that there are a lot more than 40 points in the Bush>80% category. How could that be?

[Febble]

13. I can answer your point 4:

There are four data points in the high Bush category around or above alpha=5 (actually it's nearer 4.5, around ln(1.5)), and three in the high Kerry category - but two of the Kerry ones are higher. The next highest Bush one is matched by four at the Kerry end.

There are 40 data points in the high Bush category - they are even countable if you are patient, or you can trust me. And the linear slope is not significant (p=.230), although it is positive (r=.034). But as Bruce points out in his paper, the positivity of the slope, insignificant though it is, is actually leveraged by the low points at the Kerry end, not the high points at the Bush end. Are you sure you are looking at the WPE_index plot, not the WPE one?

[Time for change]

14. Yes, I'm almost posive

However, I went back to try to look at it, and I can't pull it up.

[Febble]

15. The links are here:

http://www.dailykos.com/story/2005/5/24/213011/565

I won't link directly because I can't figure out how to scale the image. And if you want to count dots you'll want the full Monty anyway.

[Time for change]

40. Ok thanks, I don't know what the problem was

Maybe I was looking at a scatterplot with a different scale. It couldn't have been a WPE plot, because I quoted the p value of .23.

But what about my point # 3? A Kerry response rate in the 20 Bush stronghold precincts with the most negative WPE of 83%. And, though it's not mentioned in this particular post, a Kerry response rate of 75% for all 40 Bush stronghold precincts (combined, that is). You don't find that at least a little bit implausible?

[Febble]

49. Two problems

with algebraic calculation of the response rates:

The first is that using the aggregate completion rates and aggregate WPEs doesn't give you very useful answers - because you don't know which completion rates belonged to which precincts.

The second problem is this problem of sampling rate versus completion rates ("completion rate" data is what seems to be being referred to as "response rate"). The WPE_index was derived using my formula, from the WPE and vote proportions for each precinct. It could also be derived, and give identical results, directly from the sampling rate for each precinct (sampling rate = responses tallied for a candidate/votes counted for that candidate). Bias =ln(K/B) where K and B are the two sampling rates.

The completion rates, on the other hand, are the proportion of those tallied as completing the interview, as a proportion of the total completing, refusing or being "missed". If the completion rates were directly proportional to the sampling rate, which they should be if sampling was completely random, you ought to be able to derive the completion rates from the WPE and the vote-count margin using your algebra.

However, any inaccuracy in the reporting of refusals and misses, or any bias introduced by non-random sampling will mean that the sampling rate for each candidate (tallies for that candidate/votes for that candidate) may not be proportional to the completion rates. If they are not, you will certainly get "implausible" differential completion rates, even if you do the algebra at precinct level. But the implausibility will not necessarily mean fraud. It might, but it might not.

It might simply mean that misses and refusals were under-reported. It might also mean that the "bias" operated not only at the level of refusals and misses but at the level of the random sampling itself. There is no reason to suppose this did not occur. The E-M hypothesis was simply that Kerry voters "participated" in the poll at a higher rate than Bush voters. This may have been because they refused less often, that they were "missed" less often, or that they were simply approached more often by the interviewer. Or it could have been that their votes were lost or switched.

The fact that the refusal rate data has so much variance, and that refusal rates were not higher in Bush precincts suggests that IF bias arose from differential sampling rates, it operated at least partially at the level of sampling itself.

This is not inherently implausible (at least to anyone who has ever done a street survey), even if it is not what happened.

But the bottom line is: differential sampling rates (as measured by alpha) have very large variance - in both directions. If alphas above a certain level are going to be interpreted as fraud, it seems only reasonable to regard alphas with equivalent reciprocal value with the same suspicion - and there are quite a few. It is only the absence of any very low alphas in the high Bush category that lets the mean WPE rise above the level of the moderate and even categories, not the high alphas in that category per se. And one low alpha misses by an expensive whisker (expensive in terms of the time spent on this debate!)

[Time for change]

51. I'm confused

By the definitions you refer to I should have been talking about 'completion rates' rather than 'sampling rates'

The way I understand it is this: There are three categories we can refer to: responses, refusals and misses, right? Completion rate = responses divided by all three categories combined. So how does sampling rate fit into this scheme, by the definition you note above? In other words, what is sampling rate in terms of responses, refusals and misses? You say "votes counted". So what's that? Don't all three categories belong to the "votes counted" category?

[Febble]

56. The E-M report

postulated that hypothetical participation rates of 56% for Kerry voters and 50% Bush voters would explain the data. This was interpreted in some quarters as implying a uniform completion rate of these values for each set of voters. I have to admit, I always regarded it as giving an estimate of the overall ratio between two estimates of participation rate, however defined. However, the fact that it was expressed as two rates, rather than as a ratio, implied that it was being related to the completion rates, which, from the tables given in the report, appear to have averaged at around 53%. And it also told us that mean completion rates as a measure of how willing people were to participate in the poll when asked reflected a fairly low level of enthusiasm.

Completion rates, it would appear from the E-M report, are calculated as: 100% minus (refusals + misses), and tables of completion rates for each category of precinct by partisanship are given in the report, together with the refusal rates and miss rates from which they seem to be derived.

These completion rates were treated, fairly, as "response rates" in the original USCV report, and it was the slight rise in "completion rates" (stated in the E-M report as non-significant) that were cited in the USCV report as evidence that participation was not, as would be implied by the hypothesis that Bush voters participated at a lower rate, lower in Bush territory.

However, the "K/B" ratio derived from the WPE gives you the ratio between what I am calling "sampling rates". I'm defining "sampling rate" simply as the proportion of voters for each candidate that were sampled, i.e. the number of tallied responses for each candidate (so many voters saying they had voted for Kerry, so many for Bush) divided by the number of votes counted for that candidate. In some very large precincts the total sampling rate might be quite low (say 10% of all voters). In these precincts, the sampling rate for Kerry voters might be 7% and for Bush voters 5%. If the sampling procedure was completely random, you might expect a matching pair of completion rates (say 70% and 50%). But if the bias was occurring at the sampling level, rather than at the level at which refusals were recorded, the refusal rates for the two sets of voters might be a lot closer - if only those looking reasonably willing were being approached. So your completion rates might be nearer to 70%:60%.

So in this instance, the ratio between the sampling rates and the ratio between the completion rates wouldn't match. But it wouldn't mean that fraud had occurred. It would mean that either refusals weren't being properly recorded (possible) or that fewer Bush voters were being selected for interview in the first place. And in the face of generally low completion rates, it would be surprising if there was not a tendency to go for a friendly face.

I may not be using the term "sampling rate" in the way used elsewhere - I have seen all sorts of terms used to mean all sorts of things. But in terms of deriving K/B from the WPE and the vote count proportions, the ratio you get is the ratio between sampling rates as I've defined it above. If it doesn't match "K/B" where K and B are defined as completion rates, it could mean fraud, or it could mean that the bias occurred at the sampling level. So if you derive "K and B" algebraically from a known mean "completion rate" you might get very improbable answers. The reason may simply be that the completion rates were not a good measure of the actual participation rates.

Sampling bias tends to be insidious. On the other hand, so does fraud.

[Time for change]

57. Thank you for your explanation

But I still don't get it.

Please see my pm to you.

[Febble]

58. I'll have a go

at answering here too:

What I am referring to as total sampling rate is:

responses/votes

So each candidate's sampling rate would be:

responses for candidate/votes for candidate. The ratio between the sampling rates (let's call them Ks and Bs) will give you what I call alpha. Ln(Ks/Bs) will give you WPE_bias.

These two sampling rates are knowable - you just divide the tallies by the count. Or, if all you have are the WPEs and the vote count proportions you can use my fancy function to derive them from those.

Total completion rate is also knowable - total tallies divided by total selected (responses+refusals+misses). However, what we don't know is the completion rate for each candidate's voters - because we don't know how many refusers and misses were Kerry voters and how many refusers and misses were Bush voters.

We can attempt to compute it algebraically from the WPE, but that entails the assumption that the sample (including refusers and misses) was a random 1/Nth (where N is the interviewing rate) of the voters in the precinct. Where N is very small (ideally, every voter approached) the ratio between your two completion rate estimates should match the ratio betwen your two sampling rates pretty well. But where N is large, there is no guarantee that your 1/Nth of voters was truly a random 1/Nth, and in any case it will be subject to sampling error. It may have been "enriched" by voters for one particular candidate. Bush voters (or Kerry voters in some precincts) may have avoided the pollsters altogether. In precincts where refusal rates were high, interviewers may have unconsciously sampled the (N+1)th voter, if the Nth looked hostile - especially if voters were milling about. We know that WPE was greater where N was large. Refusals and misses (especially misses) may also not have been very diligently recorded. All these things happen.

All of which makes me skeptical of completion rate calculations - all we know for certain is the ratio between the sampling rates - that's what my algebra applied to Mitofsky's data gives you. It doesn't work very well (as my modelling showed) when applied to aggregate means or medians. To go on then to derive separate completion rate estimates from this ratio, given a mean completion rate for a group of precincts and some form of aggregate WPE for a group of precincts, seems to me to run the risk of piling error on error. We know the completion rate variance was large. We don't know which completion rates applied to which precincts (large or small alpha). And it is likely also that total completion rates may at least partly reflect how desperate the interviewers became for responses. After all, it was tallies, not refusals and misses, that the pollsters wanted.

[Time for change]

68. Ok, I think I got it

It seems to me then that if one attempts to calculate algebraically the K and B "completion rates" from the WPE and the total completion rate, then what we actually get is the sampling rate -- correct?

Please don't tell me no -- if you do I'll have to go through all the computations, because I'm pretty sure that's right.

If so, then what I actually calculated algebraically was the sampling rates. In other words, the sampling rates for the 40 Bush stronghold precincts were 75% for Kerry and 53% for Bush.

And if that's correct, does that sound plausible to you, in the absence of fraud?

[Time for change]

70. Correction to my last response

I did not mean to say that the sampling rate and the algebraically calculated completion rate are identical. Rather, the ratio of K:B sampling rate and K:B calculated completion rate are identical -- Correct?

[Febble]

71. Yes!

Sorry to be so slow with that. I was in error above. I should have worked this one through.

Thanks for making me do so.

It is the calculated completion rates themselves that may be nonsensical. The ratio between must be the same.

Fantastic.

[OnTheOtherHand]

96. I'm late to the party again...

Do I have right that from Febble's #58, we agree to define "sampling rate" as responses / votes? So, the Bush sampling rate would be Bush responses divided by Bush votes -- where we would have to be clear whether we refer to "official" Bush votes or "actual" Bush votes, which may not be the same. Ditto Kerry.

And "completion rate" as responses / attempted interviews (or, to put it another way, completions divided by completions plus refusals plus misses)? So, the Bush completion rate would be Bush responses divided by Bush voters attempted to be interviewed. We cannot know that last number, but we can estimate it based on assumptions; figures about a "75% Kerry response rate in the high-Bush precincts" are estimated completion rates, yes? Or, what you call calculated completion rates.

Is that right? if not, why not?

If it is right, then the ratio of sampling rates, and the ratio of calculated completion rates, should be equal. The ratio of sampling rates should equal the ratio of actual completion rates if -- well, in one case, if attempted interviews are a fixed proportion of votes. Or, approximately, if all the deviations from that assumption tend to "cancel out" statistically.

So your question in #68 is whether an alpha around 1.4 (75%/53%), entailing a 1.4 ratio of sampling rates or of "calculated completion rates," is plausible in the absence of fraud. Yes?

You didn't ask me, but since Febble didn't answer: I think an observed mean alpha around 1.4 for the 40 high-Bush precincts is plausible in the absence of fraud because alpha is very noisy in partisan precincts, especially with smaller samples. My intuitions about "noise" are informed by Bruce O'Dell's simulation (and some work of my own). A lot will depend on whether we consider the results in the 4 high-Bush high-alpha outliers plausible in the absence of fraud -- and that will probably depend more on our intuitions about exit polls than on statistical arguments.

Our intuitions about plausibility partly depend on the assumption that "calculated completion rates" are a lot like actual completion rates. I no longer trust that assumption, but I haven't formalized my dawning intuitions about specifically how wrong (or right) it is.

[Time for change]

101. It looks like we're largely on track now

Yes, Febble and I have agreed that the ratio of sampling rates and the ratio of calculated completion rates must be equal. As far as actual completion rates are concerned, well, I haven't thought about that enough yet that I would want to risk muddying up the conversation at this time -- but that must be a very difficult question, because it seems to me that if we knew that, then we'd probably know who won the election, right?

The one remaining highly related issue that Febble and I might disagree on is whether or not the individual K and B calculated completion rates are as trustworthy as the sampling rates. I think that we agree that IF the number of total "attempts" is known, then the calculated individual K and B completion rates must be exactly as trustworthy as the sampling rates. However, Febble believes that knowing the number of attempts is problematical, and therefore that the individual calculated completion rates are not as trustworthy as the sampling rates. I don't know enough about the process to argue that. Yet, I tend to disagree with her categorization of individual K and B calculated completion rates as "non-sensical".

With regard to the assumptions I made in making my calculations, all I needed to assume was: 1) A Bush vote of 85% in the 40 Bush stronghold precincts (I didn't have the actual number, so I had to assume something); 2) The data Mitofsky presented in his report is accurate; 3) No fraud. That's it, unless I'm forgetting something -- which I doubt.

With regard to your discussion about the plausibility of the alpha of 1.4 (actually, I calculated 1.42 -- I think we need every .02% we can honestly claim), the one thing that I can't relate to is your reference to small sample size. None of my calculations have to do with individual precincts. The calculated completion rates came from consideration of all 40 precincts together, as a group -- so I think that all discussion of these figures should relate to the group as a whole. Not that I'm discounting the possibility that analysis of individual precincts can't be informative. I'm just saying that the alpha or K and B calculated completion rates shouldn't be discounted on the basis of "small sample size".

So the last issue, and perhaps the one associated with the biggest disagreement between us, as well as the most important one, is what to make of the plausibility of the 1.42 alpha (or 75% Kerry completion rate) for the 40 precincts. Now we're getting into something that might be better argued on a non-statistical basis. And here is what I recently said to Febble about that:

rBr is virtually synonymous with the assertion that exit poll bias accounted for the discrepancy between the exit polls and the official vote count. Well, how did we get to figuring out alpha for the Bush strongholds in the first place? rBr would suggest a relatively low completion rate in the Bush strongholds. Well, we don't see that, so we go to modified rBr, shy Bush voter in mixed company. So that would predict relatively high poll bias towards Kerry in the non-Bush stronghold precincts. Well, the WPEs showed just the opposite, so we calculate alpha for the five partisanship categories, and we see that alpha is disproportionately high in the last one -- very disproportionately high.

In fact, the Bush stronghold category is the only one with an alpha which is much above one. So this group is driving much of the red shift, regardless of what the slope of the line is. What does that have to do with rBr? And more generally, how does one explain that fact? I certainly don't think that the fact of a non-zero slope is reason to ignore this finding. And it doesn't seem plausible to me.

[Febble]

107. I really appreciate this conversation

and you've certainly helped me identify where I think the source of the differences between us are: thanks.

In one of my stats textbooks, there was a silly example about the dangers of aggregate data: a water skier on a cold day is aiming for the shore, and doesn't want to get off too early, and land in the cold water, nor does he want to get off too late and land on the rocks. He has two goes. The first time, he gets off 60 seconds too early, and has to swim for shore; the second time he gets off 60 seconds too late and smashes into the rocks. But, hey, he has a perfect average.

Your mean alpha of 1.42 corresponds to a ln(alpha) of .35. Looking at the plot, there are indeed a handful of precincts at about that level - about 4. There are also four that are very much higher, at about 1.6 (alpha = around 5. And there are also quite a few that are lower. We do not know what the completion rates were for each of these precincts. We know what the average was, but we also know that the variance was huge. For all we know the completion rates could have been very low in the high alpha precincts, and very high in the low alpha precincts - or the other way round - or they could have a more random relationship. So if you worked out the completion rates for each group of voters in each precinct, even assuming that they were accurate, you might get a huge range of response rates for your Kerry voters or you might get a small range. And your mean Kerry voter response rates, collapsed across the group, could be anything.

But perhaps just as important is the arbitrary nature of the category boundary. If you look at the plot again, you will see that yes, the mean ln(alpha) in that category is high. But why is it high? Is it because it has lots of high points? Is it because its high points are higher than elsewhere? Or is it because it doesn't have any low points? In fact the first and last are true. It has four high points and no low points. Add in the low point that only just misses and your mean alpha will plummet. So what this means, if we are inferring fraud from that high alpha, is:

We think fraud occurred because a precinct with very low alpha only got 79% of votes for Bush, not 80%.

We think fraud occurred because four precincts in the high Bush category had alphas of around 5, even though around 15 precincts had alphas of around that level or higher in more pro-Kerry precincts, including 3, I think, in the high Kerry precincts.

Which is a reductio ad absurdum, I know, but it seems to illustrate to me why it is necessary to stand back from those means and look at the scatter. What is happening? Where are the high alphas? Where are the low alphas? Is the mean alpha being leveraged by the extreme points, or is it evident in the more moderate alphas as well?

And does any of this mean fraud?

I don't know the answer. I do think the variance is exaggerated in the extreme precincts, and OTOH is figuring out why. I also think the variance may become somewhat skewed at the extremes, which might explain those very low alphas at the extreme high Kerry end. After all, if there was only one Bush voter, and you got him, you'd get a 100% sampling rate from your Bush voters. If you were sampling every 5th voter, you might only get a 20% sampling rate from all the Kerry voters. On the other hand if you missed him, you'd have an infinitely high alpha, so the precinct would be off the chart.

But the bottom line is that aggregate category data is potentially very misleading, both because it is aggregated, and because of the arbitrary nature of the categories. Oh, and because of the unequal Ns.

I think the problem is related to Simpson's paradox, BTW.

http://plato.stanford.edu/entries/paradox-simpson/

But thanks anyway - I'm so glad you helped me think some of this stuff through!

Maybe you'll convince me yet....

[Time for change]

110. Well thank you, so do I

Yes, our conversations, as well as your sharing your paper with me and helping me to understand it certainly have improved my understanding of this whole mess.

I do have some counter-arguments that I would appreciate your consideration of:

First, I don't think that the example of the diver is terribly relevant to this issue because, whereas a diver who dives too far into the ocean can't make up for that by diving into the rocks on his next try, a Presidential candidate can indeed make up for a poor showing in some precincts by doing much better in others.

I also think that your emphasis on extreme alphas or unequal Ns is not as relevant as you suggest. I'll consider a really extreme case (but just looking at Kerry voters to make it simple). Suppose you looked at 10 precincts, and nine of those precincts had only one Kerry voter, and the sampling rate was 100% for the Kerry voters in all 9 precincts. But the 10th precinct had 90 Kerry voters, and the sampling rate was only 10% (9 of 90 Kerry voters sampled). The nine extreme precincts would not do much to raise the sampling rate. Rather, the sampling rate for the whole group would be 19% -- much closer to the one precinct with a good number of Kerry voters than to the extreme precincts. So I just don't see that extreme values that are based on small sample sizes in individual precincts make much difference to the analysis of the sampling rate or alpha for the group as a whole. In fact, that is one good argument (extreme values based on small sample sizes, that is) for taking an alpha computed for a whole group more seriously than a scatterplot. (In fact, it just occurred to me that that could be one reason for why a scatterplot would be misleading, and perhaps for causing a non-significant slope when plotting precinct partisanship vs. ln(alpha), right?).

With regard to your objection about the boundary lines, I certainly agree that the 20% cut-offs are arbitrary, and also that analysis using arbitrary cut-offs such as these are not ideal. But the only data I had to work with used the 20% cut-off limits, so that's what I used. Anyhow, it's hard to imagine that adding a single precinct to the analysis could make alpha plummet, as you say. Do you have the data to do that analysis - i.e., to add the next highest Bush vote count precinct to the data and re-calculate it for alpha? If you do, I'd be interested to know what you find.

Also, I have to say that part of the reason that I find the alpha of 1.42 (or 75% Kerry voter completion rate) so suspicious for fraud is the confluence of all sorts of other factors into the situation, such as the use of voting machines programmed with secret software. I was taught during my epidemiology training to look not only at the raw numbers but also to try to explain why they are the way they are.
When you have trouble doing that in the context of a given hypothesis (e.g. rBr), then the plausibility of that hypothesis should be thrown into question. I find a 75% completion rate for Kerry voters, compared with a 53% completion rate for Bush voters to be not very conducive the the rBr hypothesis and therefore not very plausible.

More importantly, one thing that we definitely do agree on, which is more important than all the above, is that regression analysis needs to be done on the whole data set.

[Name removed]

112. Deleted message

Message removed by moderator. Click here to review the message board rules.

[OnTheOtherHand]

126. on the small-N thing

Suppose you looked at 10 precincts, and nine of those precincts had only one Kerry voter, and the sampling rate was 100% for the Kerry voters in all 9 precincts. But the 10th precinct had 90 Kerry voters, and the sampling rate was only 10% (9 of 90 Kerry voters sampled). The nine extreme precincts would not do much to raise the sampling rate. Rather, the sampling rate for the whole group would be 19% -- much closer to the one precinct with a good number of Kerry voters than to the extreme precincts.

I'm not sure I've wrapped my head around your example here. But as far as I can think, every statistic reported by E/M would be calculated per precinct, then summed and divided by the number of precincts. That may not be true for the response rates they report, but I'm almost positive that it is true for the WPEs on which all the calculated completion rates (as far as I know) have been based. Hence the emphasis on extreme alphas and unequal Ns.

I will read over the rest of this and say more when I can.

[Time for change]

133. Here's what I think

1. The WPEs presented in Mitofsky's report, and on which I based my calculations of completion rates (75% Kerry, 53% Bush in the 40 Bush stronghold precincts) is NOT calculated by precinct, then summed and divided by the number of precincts, as you surmise. If I am correct about that, then my example is valid, and the issue of extreme precincts is not relevant to my calculation.

2. The scatterplot and the regression analysis that looked at the slope IS based on individual precinct level data. If I am correct about that, then the regression analysis is subject to a great amount of variance because of extreme values in small precincts.

3. If I'm right about both of the above, then the very high Kerry calculated completion rate and alpha for the 40 Bush stronghold precincts (which is not anywhere close to being the case for any of the other partisanship categories) have a significance that should not be diminished by the fact that the slope of the regression line is not significantly above zero. The slope is positive, but due to the large amount of variance it does not reach statistical significance.

4. It should be easy to find out if I'm right about the above. I would assume that Febble knows.

[OnTheOtherHand]

164. WPEs in Mitofsky's report

I was really sleepy last night, but I think the answer is clear. The reported WPEs must be calculated by precinct. Otherwise what could it mean to calculate a mean and median for each category of partisanship?

Based on simulation work, I imagine that the precinct median is close to the true weighted mean, but I would not bet too much on that surmise.

I agree that we shouldn't discount the 40 high-Bush precincts simply on the basis that they don't induce statistical significance in the regression line. I won't try to loop back through all the other arguments about those precincts right now.

[Time for change]

172. Well, if you're right about how WPE was reported

Then I agree that my calculations on the wide disparity in calculated completion rates between Kerry and Bush voters in the Bush strongholds has much less meaning than I thought, and in fact I would retract those calculations. Also, I think that that would be a very poor way to report "mean WPE".

What would you say if I'm right?

[OnTheOtherHand]

174. first you'd have to tell me

what "median WPE" meant in your interpretation. One of the most striking things about the 40 high-Bush precincts is the wide gap between mean WPE and median WPE, so I need to be able to interpret it somehow. But if we assumed that over the perhaps 1600-2000 people (I think the average completed N per precinct overall was somewhere between 40 and 50) in the high-Bush precincts, they indicated say 20% for Kerry when the true population figure was 15%, I too would find that hard to account for as response bias alone.

I certainly agree that "mean WPE" calculated as I describe is close to being a garbage stat. That was pretty much the point of Febble's paper, I thought -- obviously it's about WPE itself, but also about why it is susceptible to skew.

If we have been disagreeing all this time about the definition of mean WPE without realizing it, then much makes more sense.

(Edited because I originally put the parenthetical in brackets and it was interpreted as a request to italicize!)

[Time for change]

176. Median WPE would have to be calculated as you describe it

I think it's obvious that the source of at least this disagreement was based on a different understanding of the definition of mean WPE, because if your assumptions are correct then I agree with you.

It never even occurred to me that Mitofsky would have calculated it like that, and that's why I couldn't understand your arguments. In epidemiology, we never calculate mean values for a population like that. That would give very misleading results, because it would give disproportionate weight to the small units used in the analysis.

[Febble]

173. What I am assuming you have done

is to calculate the two completion rates for a notional precinct that has a with WPE equal to the mean WPE for the high Bush category; a Bush percent of vote equal to some value between 80% and 100%, and a completion rate equal to the completion rate in high Bush category.

Is this correct? Or have I got the wrong end of the stick again? If this is what you did, I am sure you did the algebra correctly - I'm just not clear what your inputs were.

If you can confirm or put me right on the above, I'll take it from there!

Lizzie

[Time for change]

175. What I did was

calculate the Kerry and Bush completion rates for the 40 Bush stronghold precincts combined, based on the equations and definitions that we have previously discussed. I did this based on the assumption that the mean WPE presented in Mitofsky's tables was calculated by the formula: WPE = K(v) - K(p) -B(v) - B(p), where K(v) and the other three terms are based on the entire group of 40 precincts -- NOT, as OTOH surmises an assumption that mean WPE was calculated by calculating the WPE separately for each of the 40 precincts and then dividing by 40.

If OTOH's assumption is correct, then your and his statements about the limitations of my finding is correct, and I agree that they have speculative value. But if my assumption is correct, then your statements about the limitations of my calculations based on extreme outliers is incorrect.

The question to you is: Which assumption is correct, mine or his?

[Febble]

177. Sadly, OTOH is correct!

I'm so sorry I didn't spot this before.

It took me ages to figure out what the WPE really was, it seemed like such a daft measure (and the description in the E-M report is not very clear, to say the least). Eventually I looked it up in a text book. It really is as simple as the difference between the margin in the poll for a precinct and the margin in the count, expressed in percentage points.

So immediately I realised, from signal detection theory, in fact, that this was not valid! So as an epidemiologist presumably you had the same doubts. But that is really what it is.

[Time for change]

180. I'm not sure that you understand my question

Are you saying that the mean WPE for each partisanship category was calculated by separately calculating the WPE for each precinct, adding the WPE for each precinct, and then dividing by the number of precincts in the category?

[Time for change]

185. Well then, I guess we'll just have to wait for some more data release

before we can say anything more substantive about this.

But I would also like to suggest that, although your bias index makes up for some of the problems with WPE (i.e., the confounding by precinct partisanship), it doesn't make up for the problem of extreme outliers with small sample sizes exerting inappropriate influence on the slope of the regression line.

I'm not suggesting that there is a better way to calculate the slope of the line -- just that extreme outliers with small sample sizes pose a problem.

And I also think that the within precinct category error should be calculated and used to calculate the K and B completion rates (and alpha and ln(alpha) for each partisanship category because, whereas the calculations that I noted previously are not as precise as I once thought, they are still suggestive, and I think that the information would be very illuminating.

[OnTheOtherHand]

213. I agree

The outliers are a nightmare, not solved by the bias index. And more information would be illuminating. Bear with me, I hope to have a few more interesting things to say along the same lines, but I first need to catch up with backed-up work from family commitments.

[Febble]

178. Sorry, replied to my comment instead of yours

reply is number 177. Also emailed you.

[Febble]

179. delete

something seems to be wrong with the thread order

[tommcintyre]

134. Hey OTOH, isn't this REALLY "Faith-Based" statistical analysis?

Ultimately, isn't this ALL ya got to work with?

1 -You have NO direct evidence to work with (since the non-responders are anonymous, and can't EVER be asked WHY they didn't respond).

2- And even your inferred data that you have to work with is so fatally incomplete (weak); since the NEP and/or Mitofsky won't release the "raw data" (and even if they did, HOW would we ever know it wasn't cooked?).

So, considering ALL of this, isn't it really IMPOSSIBLE for you to EVER have anything more than an "academic" discussion (i.e. of no PRACTICAL value in EVER determining why the non-responders DIDN'T respond)?
<Of course, the exit-poll-discrepancies-point-to-election-fraud argument doesn't really need the statistical analysis since HISTORICALLY, that has been WELL VERIFIED, right?>

Isn't this why y'all continue to have these seemingly endless unproductive discussions about, in effect, how many statistically inferred (and further obscured by hopelessly incomplete data) "angels can dance on the head of a pin"?

If I'm mistaken on any of this, please let me know.

But I have two requests:

1 - Please address ALL the ideas above that you either completely, or even partially disagree with (otherwise, I can assume you are in agreement with that particular idea?).

2 - Please explain it in VERY PLAIN ENGLISH (no WPE'S, etc.) so EVERYONE that reads here can understand.

Thanks very much, I'm looking forward to your response.

[Febble]

146. Some responses:

1: absolutely right. Everything is inferential.

2: absolutely right. Only a multiple regression can address the issue and a) if it was done before, the details were not made public, and b) the original analyses were done using a measure that was flawed. I think mine is better.

I think the value of the debate is that while it is relatively easy to know what to look for if your hypothesis is that sampling bias was responsible for the discrepancy, it is much harder to know what to look for if you are looking for evidence of fraud. Much of the debate on this forum has helped develop testable fraud hypotheses, as has Bruce's work. So far the tests have come up negative, but that does not rule out others. Several posters, including Time For Change have suggested alternative patterns that might be indicative of fraud.

But I do disagree with you that because you can't measure an effect directly (responder bias; fraud) you can't make inferences from the data, given good hypotheses. Most things we measure in science are measured by proxy. A thermometer doesn't measure the temperature - it measures the expansion of the mercury. We have to assume that the mercury expands because it is hot. It might be fraudulent mercury. If we think the exit poll discrepancy might be due to sampling bias, and the discrepancy is greater where sampling was more difficult, that is supportive evidence for the sampling bias hypothesis.

Lastly, I agree that independent analysis would be preferable to an internal analysis, even if released with full details.

But even an independent analysis will involve trusting the numbers. If you don't trust those not to have been cooked, then we might as well not be talking about exit polls at all.

[Time for change]

171. Febble, could you please see my post # 133?

We need your help on this.

[OnTheOtherHand]

165. Tom, thanks for questions about evidence

It's true that none of us has direct evidence to work with. Some people nonetheless think that the exit polls prove fraud. I would not want to say that they are the ones practicing faith-based statistical analysis, but I don't understand the source of their confidence. If you agree that no matter how much we argue about the exit polls, they won't prove fraud, then we can move on.

Has it been "WELL VERIFIED" that exit poll discrepancies point to election fraud? Maybe some discrepancies do. I don't think it's been well verified that there was massive fraud in 1992, when the discrepancies were large, but not in 2000, when the discrepancies were smaller. I haven't checked this, but supposedly the exit polls substantially understated David Duke's vote share in 1990, and I don't think it has been well verified that he stole votes. I don't think anyone has collected all the cases of substantial exit poll error and then tried to verify massive fraud in each one. If you can do that, I will be happy to see the work.

Did I miss anything big?

[LightningFlash]

271. "Regurgitated"

"Did I miss anything big?"

Maybe the fact that your own state exit polls from exitpollz.net, show an unaccounted for discrepancy in the state of Florida, which has been backed up by vote manipulation and scanned papers? :eyes:

http://www.votersunite.org/info/content/newmessup-17.asp
http://www.bbvforums.org/forums/messages/2197/6627.html?1118352564

[tommcintyre]

111. "Maybe you'll convince me yet...." YEAH RIGHT! LOL!

Maybe if you keep "throwing him bones" like this, he'll keep posting threads like this so you can keep promoting your "position" of casting doubt on the exit poll descrepency evidence?

[PATRICK]

344. I'm still in the intuition stage

where the math of any side does not weigh as much as Mitofsky seemingly cooking up a shadowy plausibility for an absolute certification that Bush won, the old cart before the horse with extreme prejudice.

But then like good, honest and brilliant thinkers we sweat out the new dance steps as dictated, because the intuiting of the truth that strikes with almost palpable force the politically minded as well as the mathematically objective that something is not only rotten in the Ukraine but in the USA, Nov. 4, 2004 as well.

That intuition is not sour grapes but a welter of airily dismissed machine failures benefiting Bush, incredibly diverse and unstopped vote suppression, unaccountability from top to bottom with no evidence or access to evidence.

So in this climate it would indeed be wise of the patient people as suspecting wrongdoing to try to get those ducks, many invisible and impossible to use as direct evidence, in order for our own benefit, not the devil's advocacy of Mitofsky which alone is treated as Gospel beyond question by the MSM.

People may worry about the squabbling that clouds any effort to get a summation out to the people but frankly it isn't just the dunderheads of the lapdog ignoramus MSM that is stupefied by the scientific details. If someone can bridge the two sides(Bushwon rationalizers, Kerrywasrobbed observers of election dirty work) without losing the mainstream in the murk of numbers and lost ballots, that would be an accomplishment indeed.

I think in conjunction with more proof of chicanery in Ohio that we need not fix the facts around an agenda, but dismiss light explanations about the uncooperative Bush voter with some open objectivity and honest debate compared to the violent dismissal of doubt itself by the side with the most to fear from open minds and the truth. THEY consider reflective, good science to be weakness and certainty voiced loudly to be the truth. And they are always making the first moves in the game they made. Suddenly everyone finds themselves shouting about the debate game rather the truth.

[autorank]

12. How about some comments on an ELECTION FRAUD thread?

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374165

The silence has been stunning. The normal debate has abated. Let's see what the comments are on this. OTOH has made the point that he likes to be "a doubter" and others have played that role as well. Rather than get hung up on internal politics, why not take on this interesting analysis?

Just asking...

[RonB]

16. USCV's Analysis is not Based on WPE, "Aggregate" Analysis is not "Flawed"

Bruce,

You say:

"The key argument of the USCV Working Paper is that Edison/Mitofsky’s exit poll data cannot be explained without either (1) highly improbable patterns of exit poll participation between Kerry and Bush supporters that vary significantly depending on the partisanship of the precinct in a way that is impossible to explain, or (2) vote fraud. Since they rule out the first explanation, the authors of the Working Paper believe they have made the case that widespread vote fraud must have actually occurred."

I believe what we say is that varying patterns of partisan exit poll response require "further explanation" beyond a "pervasive bias" theory that is somehow inherent to being a Bush voter in that it is independent of precinct partisanship - our interpretation at the time of "rBr". The effort as I see it is to move the analysis beyond these vague and general "explanations" toward more serious substantive explanations that relate to the actual factors that influence response rate.

I don't believe we ever state that implausible invariant mean bias (K/B) is sufficient evidence of vote fraud. Just that a more substantive (so far not offered) explanation is required.

However, a closer look at the data they cite in their report reveals that Kerry and Bush supporter exit poll response rates actually did not vary significantly by precinct partisanship.

My analysis (for calculations see previous post - summary table copied below) demonstrates conclusively (I think) that K/B levels are indeed different at a 5% two-tail level of statistical significance for certain partisan categories of precincts. As you can see, it is particularly hard to see how the K levels calculated from both mean, and median (to eliminate out lier influence) WPE's for high Bush precincts could be obtained in any of the other categories of precincts.

These 95% confidence intervals are based on random sample exit poll selection - the same methodology use to analyze exit polls - without any cluster adjustment as there is no bias in estimating single precincts - and with mean approximates of R and b and k levels. If these later are not significantly correlated with each other (and the data suggest that they are not) these estimates should be quite accurate.

Means and Medians - and not individual data points, are after all, what statistical inference is usually about. Simulation in many ways is a "backward" reconstruction exercise, that cannot really add much (except for the additional hypotheses used to generate the simulations) to the original mean and median "information" about the data.


Table 2: Mean Based Calculations:

95% Confidence Interval for B 95% Confidence Interval for K
Covers all B? -1.96 x SD' B +1.96 x SD' Covers all K? -1.96 x SD' K +1.96 x SD'
Yes 20.9% 53.8% 86.6% No 49.3% 52.9% 56.6%
Yes 36.8% 49.6% 62.4% No 51.8% 57.3% 62.8%
No 43.2% 47.6% 52.0% No 52.0% 56.4% 60.8%
Yes 49.1% 52.6% 56.1% No 52.5% 60.6% 68.7%
Yes 47.1% 52.9% 58.7% Yes 31.9% 84.0% 136.1%


Table 4: Median Based Calculations:

95% Confidence Interval for B 95% Confidence Interval for K
Covers all B? -1.96 x SD' B +1.96 x SD' Covers all K? -1.96 x SD' K +1.96 x SD'
Yes 19.1% 51.9% 84.8% No 49.5% 53.1% 56.8%
Yes 37.1% 50.0% 62.8% No 51.7% 57.2% 62.7%
No 43.3% 47.7% 52.1% No 51.9% 56.3% 60.7%
Yes 49.1% 52.6% 56.1% No 52.5% 60.6% 68.7%
Yes 48.4% 54.2% 60.0% Yes 20.2% 72.2% 124.3%



You say:

"Systematic exit poll bias cannot be ruled out as an explanation of the 2004 Presidential exit poll discrepancy – nor can widespread vote count corruption. The case for fraud is still unproven, and I believe will never be able to be pr oven through exit poll analysis alone."

I agree.


You say:

"The fact that I chose not to endorse the USCV Working Paper should be a clear indication that I do not support its central thesis, and in fact believe that the simulation data they cite refutes the Working Paper’s conclusions.

I am not a statistician, but as a computer systems architect, I create mathematical models to simulate the performance of large-scale computer systems, and mathematical simulation of the cost and efficiency of business processes is a significant part of my consulting practice. My own election simulation results are cited on pp. 9 -10 and in Appendix G of the May 12th Working Paper; as the creator of the only USCV simulation which accurately reproduces aggregate Mean WPE, Median WPE and participation rate data from the E/M January report, I feel an obligation to ensure that my work is correctly interpreted."

Yes, definitely. The O'Dell simulation is an "output" simulation whose purpose is to simulate the E-M outcomes. The Dopp simulation is an "input" simulation whose purpose was to simulate the E-M hypothesis (or "hypothetical") inputs.


You say:

"I can show that several of the USCV election simulation programs are flawed, and that when the Liddle Bias Index is applied to the “USCV O’Dell simulation” data cited in the Working Paper, it produces results consistent with those recently reported by Warren Mitofsky for the E/M data as a whole."

I don't know what you mean by "flawed" but if you mean they don't reproduce the E-M outcomes thats the point! (see previous comment).

The Dopp simulator, in particular, as interpreted in Appendix I shows that the E-M outcomes are not obtainable under (what we then thought was) the E-M constant invariant K=.56, B=.5 hypothesis.


You say:

"I respect Ron's opinion, but his insistence on using aggregate WPE as a tool to interpret poll response bias (or vote fraud) is mistaken."

I think you're over stating your case here. There is nothing "inherently wrong" with "aggregate analysis", i.e. using means and medians. All these different types of analysis serve different purposes. Simulation adds the "information" (beyond the means and medians) that it assumes about the distributions to generate the simulations (see previous comments). This can be useful if there is some plausible "additional" information about these distributions and if there is a concern that "aggregation bias" resulting from parameters being correlated with each other (see previous post) will affect the analysis.

Using aggregate statistical analysis (as I do in the Tables above) is particularly useful to get significant probabilities, something that is particularly hard with simulations as there are so many possible options for distributions etc.


You say:

"His analysis of the Liddle Bias Index is also off-target. Liddle's Bias Index is an inherently superior metric to WPE, and analysis based on aggregate WPE are highly misleading."

I've stated all along that Liddle made an important contribution (that we acknowledged in the report) by focusing on looking at the "overall pattern" of K/B over partisan precincts. USCV had previously focused on calculating K and B in precinct quintiles and looking at the plausibility of these values and their ratios and differences across precinct quintiles, rather than at the overall pattern across precinct partisanship.

However, the basic point is that the calculation of and focus on K and B (the "confounded partisan response rates) was initiated by USCV. Liddle's alpha is simply LN(K/B). She derived it without K and B, apparently independently - its not a big deal to derive any of these formulas - the insight is in the need to do it and how they can be used. Liddle's contribution is in the later.


Unfortunately Liddle chose to frame her analysis as a new breakthrough in using "confounded variables" rather than simply a new way to use K and B to do the analysis. If she had been more clear about where her ideas came from, or at least who had done them before, she would have had to explain how she came to opposite conclusions from the same variables. I don't believe she did any of this deliberately but it has been the source of endless confusion that I think, most unfortunately, has gotten you confused as well.

NOBODY IS BASING ANY ANALYSIS ON WPE. ALL OF US ARE USING K AND B. WPE IS NECESSARY TO DERIVE K AND B BUT THE ANALYSIS IS BASED ON K AND B VALUES NOT ON WPE. WE ARE ALL ANALYZING THE SAME VARIABLES IN SLIGHTLY DIFFERENT FORMS (K, B, K/B, K-B. LN(K/B))

Liddle's other contribution was to point us in the direction of simulations. This is another method of analysis that could show the impact of "aggregation" effects. We used this in the report to show that these aggregation effects to not vitiate the conclusions of the aggregate analysis.

Ultimately this debate cannot be resolved with absolute certainty unless Mitofsky gives us the actual data (or does the calculations form the actual data himself). However, short of this I see no reason to make claims that simulations constitute absolute proof and "aggregate" statistical analysis is "off target" or fundamentally wrong.

The aggregate results show, with high probability, that K/B is not constant across precinct category. Your simulations show that, I guess, that this might not be the case. However, I don't think you (like Liddle) should jump to the conclusion that K/B is unvarying. It seems to me that our report (and the analysis above) makes a very plausible case that "more stuff needs to be explained" - and this is the point of it all!


Best,

Ron

[Febble]

21. Some points

I agree we are all using K and B. What I proposed was that that ln(K/B) (which I called log(alpha) was a better dependent measure than WPE. So that in any multiple regression analysis - and I totally agree with you that multiple regression is what is required - I believe that ln(K/B) would be a better measure than the WPE used by E-M in their report.

And of course I derived ln(K/B) from K and B, and indeed both appear in the formula (different letters, but whatever...) You couldn't derive any formula to measure differential sampling rate without variables representing the two differential sampling rates. However, without the data we don't know K and B so I just did the algebra to derive, not K and B, which you did but "alpha" (K/B) which doesn't depend on the absolute value of either. Different horses for different courses. I wanted a dependent measure for use in multiple regression. You wanted K and B.

And enough already with the "apparently independently". The algebra was no biggie, I agree. And the reason I billed it as an answer to a "confounded variable" was nothing to do with your analysis, but with E-M's analysis - they one they had done, and the one which IMO they needed to do, which was to run a multiple regression with ln(alpha) as the dependent variable. In statistical terms, the WPE is "confounded" as it contains a term you also want on the predictor side of the equation. Confounds don't come much more confounded than that.

Finally: I certainly have never jumped to the conclusion "that K/B is unvarying" - my point, throughout, has been that it varies far more than I thought any of us had been allowing for. You thought my alpha of 2:1 was extreme - in fact there are 2:1 alphas, as well as 1:2 alphas all over the plot. One of the extreme sillinesses of this debate has been the endlessly extending faux equation: Liddle=constant bias=rBr=Mitofsky=Karl Rove=O'Dell.....

A flat regression line between bias and vote-count margin means that bias does not vary linearly with vote-count margin. That is all. It does not mean bias is unvarying; it does not mean that it does not result from fraud; it does not mean that it does not vary in some complex and informative non-linear way with vote-count margin (though I have yet to see in what way it may). And above all, it does not prove rBr. It merely refutes "Bsmvcc". In fact I am starting to conclude that any "fraudulent" account of the Great Exit Poll Discrepancy may be more consistent with "Ksmvcc".

[Bruce ODell]

38. We part company when it comes to confidence in WPE

Back at you, Ron!

Baiman(Ron's text is in blue):I believe what we say is that varying patterns of partisan exit poll response require "further explanation" beyond a "pervasive bias" theory that is somehow inherent to being a Bush voter in that it is independent of precinct partisanship - our interpretation at the time of "rBr". The effort as I see it is to move the analysis beyond these vague and general "explanations" toward more serious substantive explanations that relate to the actual factors that influence response rate.

I don't believe we ever state that implausible invariant mean bias (K/B) is sufficient evidence of vote fraud. Just that a more substantive (so far not offered) explanation is required.

O'Dell: Ron, I'm afraid that is not quite the full story. On May 14th USCV released a press release to coincide with your trip to AAPOR that stated:

"The National Election Data Archive (NEDA) today has released a new report, demonstrating that data from the Edison/Mitofsky analysis is consistent with the hypothesis of a corrupted vote count, and inconsistent with the competing idea that Bush voters were under-sampled in the poll. Using numerical modeling techniques to simulate the effect of polling bias, NEDA scientists are able to reproduce signature patterns in the Edison/Mitofsky data by incorporating a general shift in the official vote tally in the model."

"Most telling is the fact that the highest participation rates and the peak disparity between poll and official returns both occurred in precincts where Bush made his strongest showing. This feature of the data is inconsistent with the Edison/Mitofsky assumption that polling bias was responsible for the gap."

Do you still stand by those words? You must have realized that they would galvanize the voting activist community. If you no longer believe that statement to be true and accurate, isn't a formal retraction appropriate, rather than a quiet shift in tone?

At any rate, while your tone may be a little more measured now, the underlying analysis is still the same. The major thrust of the Working Paper is that the simulations and aggregate calculations require implausibly large exit poll response rates for Kerry supporters , especially in High Bush precincts. Check your own Appendix B. And clearly the argument on pp. 8-9 of the paper is an attempt to show how response bias alone cannot explain E/M's Within Precinct Error numbers. I'll address the reliability of WPE below.

Baiman: Means and Medians - and not individual data points, are after all, what statistical inference is usually about. Simulation in many ways is a "backward" reconstruction exercise, that cannot really add much (except for the additional hypotheses used to generate the simulations) to the original mean and median "information" about the data.

O'Dell: technically, doesn't my simulation and Mitofsky's data represent interval data that has actually been accurately calibrated to the nomimal Mean and Median data you prefer to model? In other words, my precinct-level simulation data (and the actual precinct level E/M data) rolls up to your aggregate Mean and Median data, but also enables additional analyses to be performed. My experiemce with simulation is that when properly calibrated, even a simple simulation model can accurately reproduce complex behaviors of the whole system. I didn't and I still don't understand your reluctance to grapple with the available interval data.

Baiman: The Dopp simulation is an "input" simulation whose purpose was to simulate the E-M hypothesis (or "hypothetical") inputs... I don't know what you mean by "flawed" but if you mean they don't reproduce the E-M outcomes thats the point!

O'Dell: there are several Dopp simulators. The one I criticized - on pp. 8-9 of the report, clearly is intended to reproduce E/M's outputs by a combination of "credible response bias" and "fraud" as the report states on p.9 : "The chart below shows how similar the vote shift + expected response rates simulation look to the actual Edison/Mitofsky data. The pattern produced by a 6% vote shift + the assumed response rates for Bush and Kerry voters given at left are shown in the above two charts. Notice that it is closer to the pattern of the actual Edison/Mitofsky data than the pattern produced by a simulation of the rBr theory of Kerry and Bush response poll bias with means 56% and 50%. The patterns can be made more alike with small response rate changes." To me this indicates the goal of the simulation on p. 8 is to match E/M's results. And it does not come close.

I did not discuss the Dopp Appendix I simulation in my paper, since Kerry +6.5% bias obviously can't match the E/M WPE signature and participation rates; criticizing Kerry +6.5% seems like beating a dead horse. Also, since WPE is inherently skewed, it is just misguided to use it as the basis of any simulation.

Baiman: There is nothing "inherently wrong" with "aggregate analysis", i.e. using means and medians.

O'Dell: my point was it was the use of aggregate WPE that was inherently flawed.

(To those of you still trudging through this exchange - rest assured, we're almost home.)

Baiman: Liddle's alpha is simply LN(K/B). She derived it without K and B, apparently independently - its not a big deal to derive any of these formulas - the insight is in the need to do it and how they can be used. Liddle's contribution is in the later.

O'Dell: it's a matter of public record that Liddle published the bias index on April 7, causing you to update the March 31 report, on April 12.

Baiman: Unfortunately Liddle chose to frame her analysis as a new breakthrough in using "confounded variables" rather than simply a new way to use K and B to do the analysis. If she had been more clear about where her ideas came from, or at least who had done them before, she would have had to explain how she came to opposite conclusions from the same variables. I don't believe she did any of this deliberately but it has been the source of endless confusion that I think, most unfortunately, has gotten you confused as well.

O'Dell: Ron, honestly, I don't think I'm confused. If you look at the chart on page 33 of my report, you can see that a single value of K/B (or alpha, or ratio of the response rates of Kerry and Bush) produces disproportionately more WPE for High Bush than for High Kerry precincts, and the higher the value of K/B, the more disproportionate it gets . That mathematical result is what that mathematical result is; it's really not a question of interpretation.

Baiman:NOBODY IS BASING ANY ANALYSIS ON WPE. ALL OF US ARE USING K AND B. WPE IS NECESSARY TO DERIVE K AND B BUT THE ANALYSIS IS BASED ON K AND B VALUES NOT ON WPE. WE ARE ALL ANALYZING THE SAME VARIABLES IN SLIGHTLY DIFFERENT FORMS (K, B, K/B, K-B. LN(K/B))

O'Dell: I just don't follow your disclaimer on WPE. The charts on pp. 4, 5, 6, 7, 8 and 9 all have WPE on the Y-axis, Appendix B has an extended discussion of absolute and signed WPE and Bsvcc, and in fact it seems as if the main thrust of the paper concerns WPE. The variables you cite are not all equally useful or interchangeable when it comes to analyzing response rates. WPE introduces an inherent skew to all calculations based on it. Your reported highly differential partisan response rates, as calculated using aggregated and non-aggerated WPE, are a mirage.

Baiman: The aggregate results show, with high probability, that K/B is not constant across precinct category. Your simulations show that, I guess, that this might not be the case.

O'Dell: in fact, my simulation data supports your thesis - when the warped yardstick of WPE is used. When a straight ruler (Liddle's Bias Index) is used on that very same data, highly variable response rates across categories of patisanship - your key finding - vanishes. So does the Bush stronghold vote fraud thesis in Appendix B page 16.

Doesn't that raise even a small doubt in your mind?

We both agree that the data is consistent with some combination of exit polling problems and vote fraud. When measured by Liddle's Bias Index, there are suspiciously biased results appearing to favor both Bush and Kerry in all categories of partisanship. For a non-partisan organization, USCV has said little about the potential for vote fraud in the High Kerry precincts, although there are certainly some anomalies worth looking into there. In fact, whatever vote fraud occurred in 2004 is going to take a much more sophisticated analysis - and accurate measuring tools - to track down.

I'm so worried about finding out how much fraud actually occurred that I've decided to make all my measurements with an unwarped yardstick and report whatever I see when I use it - especially if it confounds my preconceptions. No matter how often you tell me WPE is a great tool to use, Ron, that yardstick is still warped. Drop it and let's all use the right tool for the job.

[RonB]

380. Bruce - I'm Your Thoroughly Confused - I'm sorry this happened

O'Dell: I just don't follow your disclaimer on WPE. The charts on pp. 4, 5, 6, 7, 8 and 9 all have WPE on the Y-axis, Appendix B has an extended discussion of absolute and signed WPE and Bsvcc, and in fact it seems as if the main thrust of the paper concerns WPE. The variables you cite are not all equally useful or interchangeable when it comes to analyzing response rates. WPE introduces an inherent skew to all calculations based on it. Your reported highly differential partisan response rates, as calculated using aggregated and non-aggerated WPE, are a mirage.

Baiman:

Unfortunately, I just composed a lengthy response and lost it - but a summary:

The charts are all trying to simulate Mitofksy data and particularly the WPE patterns - the simulations are not based on WPE but rather on K and B.

USCV initiated the analysis of K and B (also a matter of "public record") in our March report and earlier.

Appendix B was produced in response to Liddle's empirically pointing out the asymmetric pattern. It shows the underlying mathematical reason for the asymmetry and shows that (even with alpha=1.15) it results in a skew of only 0.3% WPE (see Appendix F Table 2 p. 18 May report).

We are all using the same "unwarpped" yard stick. In fact USCV initiated the use of this yard stick. In fact I think we need to "unwarp" it further by taking into account random sampling error (not data variance but model variance) as this will allow disentangling of sampling error from other non-statistical sources of error.

My "aggregate analysis" has reproduced O'Dell simulation data for K and B, mostly within 1% and without the unexplained correlations and variances used in the O'Dell simulation - some mirage! (see recent post).

Its a mathematical fact that Liddle's alpha=K/B as we calculated it in previous reports.

[tommcintyre]

17. Bruce O'Dell IS CLEARLY advocating for Mitofsky's position

You state:
"...as far as I can tell, Bruce O'Dell is not advocating for Mitofsky's position..."

Bruce O'Dell says <above>:
"...it produces results consistent with those recently reported by Warren Mitofsky for the E/M data as a whole."

That's pretty clear that he is "advocating for Mitofsky's position".

And, of course, he is also trying to "create doubt" by "...merely saying that USCV did not prove him wrong...". <italics mine> That has been, and will continue to be, their modus operandi.

Furthermore, that is NOT all he "merely" attempted to do, is it? His subject line stated that the USCV paper is "fatally flawed". This is a CLEAR attempt to discredit the USCV paper, isn't it?

Time for change, I really don't know what your motives were in making this thread; but one thing I can say clearly - at a minimum, I believe you have demonstrated extremely poor judgment in doing it.

I am currently researching, and composing, a response to O'Dell in the original thread. The subject is: "Bruce, your post is "fatally flawed"". When finished, I will copy it over here, since you felt you just had to do this.

[mgr]

19. I don't think your point is accurate

What O'Dell confirms is the effect of the correction using Liddle's term at the meeting. He does not resolve that the bias is from the higher non response of Bush supporters to Kerry, or a reflection of fraud, but that possibly a different attack is needed. Febble's point about looking now at the variance is one means towards this end.

Mike

[autorank]

22. Mike, here is the essence:

"it produces results consistent with those recently reported by Warren Mitofsky for the E/M data as a whole."

There are no two ways to spin that (well Ari could but he's got a gathering in the private sector).

[Febble]

23. I think you have an odd definition of spin

it seems to be that when it does not support your position it is "spin" and when it does it is just Okely Dokely.

Is this forum for trying to find the truth, or for collecting potential evidence for a court case?

Even if the latter, doesn't make sense to test the evidence for weakness before the defence counsel does it for you?

Has no-one on DU ever heard of proportionally spaced font?

[autorank]

25. febble, you said: "Mitofsky asked me to do some further analysis"

These are your words, not mine, from your post of a couple of weeks ago.

What does this say about your current objectivity?

Just asking.

[Febble]

26. I'm not claiming objectivity

it's Bruce's objectivity that was being called into question.

But since you ask, no, I can no longer be assumed to be objective, although I hope I do my best to be objective. Non-objective analysts are pretty useless analysts.

That is why independent analysts are better.

[autorank]

27. Wouldn't working for Mitofsky make you "non objective"?

[Febble]

31. Oh no, that guy in the shades is back!

What happened to Conyers?

OK - if I had my heart's desire, I'd be an independent analyst, preparing a report for peer-review.

I hope that someone else will get their heart desire, because your country could use an independently prepared and reviewed report.

And I am aware that any research that is paid for runs the risk of lacking objectivity - and at best cannot have any claim to be objective, even if it has been done conscientiously. Drug company research is a prime example. No drug research would get done without support from drug companies, yet we all have to rely on their research. It's not good. It also means that good cheap remedies don't get researched, because no-one stands to make a buck.

And I know that maintaining independence of thought, even when you earnestly mean to, is difficult. It's why we do double blind controlled trials. And because I know these things, I know that it is perfectly possible for good conscientious people to bring unconscious bias into their research.

And it is also why I do not find biased sampling an implausible explanation for the exit poll discrepancy .

But I will nonetheless do, as I do daily as an honest scientist, my best to do an honest analysis. I trust me, even if no-one else does.

[autorank]

33. This is progress, we're talking about drugs!

Conyers may come back judging from the tone of this response, more in tune with our recent "era of good feelings."

[kobeisguilty]

37. honest analysis

I for one, and probably speak for many of DU'ers and others, warmly appreciate your analysis Febble. Your ability to present your viewpoint with little innuendo and a lot of fact while avoiding ad hominem attacks against those that disagree with you carries great weight among us that seek the truth. I'm not saying I agree with you 100% on all things, but the class by which you show your arguments is rare among this debate.

keep it up homegirl.

:)

[mgr]

28. Too black and white for me

That is a stongly manichean position. Why is it that everything associated with Mitofski is bad? So is O'Dell now bad?

The straight and narrow is becoming the path that fewer and fewer associated with USCV can follow. I think your world view needs to come to terms with what it means to come to consensus. USCV is not at consensus, so what. I am glad to see it, it means they are flesh and blood, and see things differently one from another, and they lead with their reason and their brains.

Mike

[autorank]

35. Them's fightin' words, I'm no Manichean, I'm a Gnostic

Two points to illustrate:

1) From the Gospel of Thomas:
2. Jesus said, "Those who seek should not stop seeking until they find. When they find, they will be disturbed. When they are disturbed, they will marvel, and will reign over all. " We're past "disturbed" and ready to "marvel" and "reign over all" with our findings.

2) The handling of data is also analogous to two discoveries of religious texts. The Dead Sea Scrolls were discovered and then put in the hands of "the select" who dole out access to them. Not much has been "revealed" and access is limited. The Nag Hamadi Gospels were discovered in Egypt and opened to inspection by all. There is a florid literature of and on these gospels.

Let our data free!

[mgr]

61. I'll keep that in mind in the future n/t

[tommcintyre]

39. "Too black and white"? Try clarity for a change, and less obfuscation

"Why is it that everything associated with Mitofski is bad? So is O'Dell now bad?"
I can't even begin to divine how you expected anyone to take this "leap in logic" seriously. This "jump" in "reasoning" is truly... well... fantastic would be accurate.


"...come to terms with what it means to come to consensus."

Yes... consensus is gooood! Who can be against consensus? Nice straw-man there mgr! And you proceeded to knock 'em down with great precision. ;)

A little problem... ODell has admitted that consensus was NOT possible.
"After unsuccessfully working within US Count Votes to revise or retract the Working Paper...I see no alternative but to publicly challenge the report’s methodology and conclusions."

It sure doesn't sound like consensus-building, does it?

Here's a link you may find useful. It may help you to improve your arguments. It's the best one I've found so far.

"A List Of Fallacious Arguments"
http://www.don-lindsay-archive.org/skeptic/arguments.html

[Bruce ODell]

42. Eventually, you'll read my paper, I'm sure... but still waiting....

You haven't replied to my challenge that had you actually read my paper , you would have encountered these words:

"This paper should not be misinterpreted as an argument against the likelihood of vote fraud. Quite the opposite; I believe US voting equipment and vote counting processes are severely vulnerable to systematic insider manipulation and that is a clear and present danger to our democracy . I strongly endorse the Working Paper’s call to implement Voter-Verifiable Paper Ballots and a secure audit protocol, and to compile and analyze a database of election results."

Actually, your postings about my paper draw liberally and adundantly from the "List of Fallacious Arguments" you so helpfully provided.

-Ad Hominem: you can't refute what I say. Therefore you attack me as a disruptive influence on a group of people you appear to admire.

-Argument by Laziness: if you had read my paper, you would not say that I "supported Mitofsky" (whatever that means).

-Appeal to Widespread Belief: since you and many of your associates apparently already know that vote fraud occurred, you don't need "proof".

-Argument by Emotive Language. Ah, far, far too many to choose from.

-Argument by Rhetorical Question. "Who can be against consensus?"

-Inflation of conflict. Ron and I disagree, you like Ron, therefore I am bad to disagree with Ron because I risk discrediting USCV by having the audacity to pursue the truth in public.

-Needling. For example, your suggestion that those who disagree with you should check a website and brush up on their rhetorical skills, with the clear implication that you already posess those skills in abundance.

-Changing the Subject. You can't criticize my thesis, so you criticize my approach.

and last, but not least, my favorite:

- wait for it-

Misunderstanding the Nature of Statistics.

As I said before - and you ignored - I'm so concerned that the election may have been stolen, I'm actually trying to find the evidence. If you've found the evidence, show me. If you haven't, either join me in analyzing the data - if you can - or stop questioning my integrity, simply because I insist on analyzing it properly.

[autorank]

47. What's your best evidence. Surely you're not looking for election

fraud as a deductive process. I ask this question from time to time and get very interesting informative answers from people.

I did read your paper, BTW, and this isn't about that.

What was it about the election, at the time and as you've looked into things, that made you think (preliminary evidence, let's say) that there was election fraud?

[Febble]

50. And my favorite, from that excellent list

is "excluded middle", and relevant here to the proposition that the election may been corrupt, but that the corruption cannot (yet) be concluded from the exit poll evidence.

George W Bush made the same argument when he asserted that "You're either with us or against us in the fight against terror." And it, too, was fallacious, tragically so.

Probing evidence for weakness is not the same as denying that the phenomenon it is evidence for does not exist.

If someone had probed the TANG memo forgery before it was used as evidence for Bush's AWOL, he might not be president today. There was plenty of evidence. The trouble was that the piece that was used was "fatally flawed".

[tommcintyre]

54. Hey Febble. Is THIS an example of the "excluded middle"? ;)

"...either join me in analyzing the data - if you can - or stop questioning my integrity, simply for insisting on analyzing it properly."

I sure can't find the "middle" in THAT statement above, can you?

They ARE very useful, especially if we can earn to apply them to our own arguments, huh?

I guess it's another case of cognitive dissonance (and other things), rearing it's ugly head again, don't ya think? ;)

[Febble]

55. No, that is not an example of an argument

with an excluded middle, or even an assertion. It is an exhortation.

These are arguments with excluded middles:

Bruce O'Dell does not support Baiman. Therefore he must support Mitofsky.

Febble does not support USCV. Therefore she must support rBr.

Both have been made on DU, by you. Both are false. Both derive from expressions you now claim to approve of such as "are consistent with". They reflect understanding of the limits of inferential data.

In contrast, here is the first paragraph from the abstract to the Baiman Dopp paper:

New evidence from mathematical simulations conclusively shows that any constant mean exit poll response bias hypothesis such as the "reluctant Bush responder" (rBr) hypothesis is not consistent with the pattern shown by the Edison/Mitofsky exit polling data. Other explanations are required to explain the Edison/Mitofsky pattern of exit poll discrepancies and overall response rates.

(My bold)

What Bruce is saying, I believe, and certainly what I am saying, is that the parts in bold are not tenable. That the evidence is "consistent with" - that other explanations are "not required".

Neither of us is saying that fraud not consistent with the data, we are just saying that differential sampling is also consistent with it. As yet.

It is Ron who is making the stronger inference.

And yes, I agree that cognitive dissonance is a major problem on this forum. It appears to take the form of intolerance of any analysis of any data, no matter whether it comes from a man of Bruce's integrity and dedication to the cause of electoral security, if it shows any sign of conflicting with the case that Kerry won the election.

[tommcintyre]

124. Oops! I THINK YOUR BIAS IS SHOWING AGAIN ;)

I think you are in "full-spin-mode" again; and, your "cognitive dissonance" protection mechanism in in full swing too. ;)

<You said:>
""No, that is not an example of an argument" with an excluded middle, or even an assertion. It is an exhortation."

Definition:
http://www.cogsci.princeton.edu/cgi-bin/webwn2.1?o2=&o0=1&o6=&o1=1&o5=&o4=&o3=&s=exhortation
"exhortation (a communication intended to urge or persuade the recipients to take some action)"

Well feeble, isn't it the purpose of ALL fallacious arguments "...to urge or persuade the recipients..."???
-----------------------------------
The excluded middle fallacious argument:

Hint: This one's REALLY EASY to spot - that is, IF one is OPEN to spotting it.

Almost any time you see an "either...or" "exhortation", you can pretty much count on there being PLENTY of excluded middle choices.

For example:

The "essence" of Bruce O'dell's "exhortation":
"...either join me...or stop questioning..."

or, GWB's:

"Either you're with us, or your with the terrorists..."

Now really, aren't there PLENTY of middles for both of the above statements? <IMO, cognitive dissonance would be the main barrier for someone to NOT recognize that there are middles in EITHER ODell's OR GWB's "argument". ;) >

I think, since ODell is your ally in casting doubt on the exit poll discrepancy evidence, you can't be objective about this.

-------------------
Ahhh, now on to your claims about what I have said:

"Bruce O'Dell does not support Baiman. Therefore he must support Mitofsky."

<So NOW we're in the realm of false logic? ;) First, I've got to wonder (just guess) WHERE you got this idea from. So here goes:>

I said:
"That's pretty clear that he is "advocating for Mitofsky's position <re: the rBr hypothesis>"."
<This IS quite a leap in logic (and VERY assumptive) on your part, isn't it? How did you connect "does not support Baiman", with "must support Mitofsky"? AND, are you implying that I also said Odell "support(s) Mitofsky" overall? (That wouldn't surprise me; but I certainly wasn't even remotely Saying (or thinking about) that here.>

"Febble does not support USCV. Therefore she must support rBr."
<WRONG AGAIN. I derive your support of rBr from your advocacy of Mitofsky in using "your" function/paper to support rBr. <BTW: Just because you have sent PLENTY of "mixed messages: regarding this, doesn't really change that fact, does it?>

And, finally, you said:
"...a major problem on this forum...intolerance of any analysis...shows any sign of conflicting with the case that Kerry won the election..."

That's your ACTUAL "bottom line", isn't it REALLY?

And YOU should be able to understand WHY "we" have such a problem with this: This has been the modus operandi of the attempted cover-up since the election was stolen, hasn't it? Create doubt that KERRY REALLY WON, wherever, and whenever possible.

And, creating doubt that the exit poll discrepancy evidence points to a stolen election has been in full gear ever since January, when Mitofsky re-fabricated rBr for the son (GWB - like he did for the father in '92). Pastor used this rBr as an excuse to keep USCV, and the exit poll discrepancy evidence, from being heard at the first Carter/Baker Election Reform Hearing. And, is there ANY doubt that your paper is being used to further justify USCV NOT getting to testify at the second hearing (coming up on June 30th)?

I've suggested this to you before: I think you really need to do a "reality check". Considering how you responded to my "excluded middle" question, AND what I just reminded you about, it's time to bypass your ego (and wishful thinking), before it's too late - TIME IS RUNNING OUT!

[autorank]

125. Brilliant analysis tommcintyre!!! Sure would help if we had a Mark Felt!

:thumbsup::thumbsup::thumbsup:

Nominated for Outstanding Post of This Thread

NEW LEADERS FOR A NEW DEMOCRATIC PARTY!

Contact the DNC and Give 'em Hell About NOT Acting on Election Fraud

[Febble]

145. I am not "creating" doubt

that Kerry won the election.

I simply doubt it. For the reasons I have given.

[Name removed]

147. Deleted message

Message removed by moderator. Click here to review the message board rules.

[tommcintyre]

150. *sigh* I don't even think she can "hear" you kster. :(

I think she has built a "wall" to "protect" herself from such responsibilities.

But, the funny thing is about such walls. They tend to collapse one day from all the weight of so much denial.

And when they do....

[Febble]

155. So, am killing American soldiers, am I?

Let's see:

Somewhere around half of American voters vote for a president who embarked on an illegal war in Iraq, which I utterly oppose. Because I opposed it I was seriously concerned as to whether it really was slightly more than half of American voters, or slightly less, who voted for that president.

The exit polls initially suggested to me that it was possible that massive fraud had taken place, and that perhaps Kerry really had been the American choice.

On closer examination of the exit poll evidence I find that the tool used to conduct an investigation into what caused the exit poll discrepancy was flawed. I suggest a better one. It is used. Its math is not disputed. Yet.

An initial reanalysis of the data, by Mitofsky, using my measure, suggests that one particular fraud hypothesis, advanced by USCV, is not supported. Mitofsky engages me to run some more detailed analyses. For all you know, these may support alternative fraud hypotheses.

Just how does any of that make me responsible for soldiers dying in Iraq?

Don't bother to answer. It's a rhetorical question.

[tommcintyre]

158. "...how does any of that make [YOU] responsible..."?

Read here... or are you avoiding (denying) this post too?

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#375215

[Febble]

162. You say:

People ARE dying in Iraq everyday, and WHETHER YOU LIKE/ADMIT IT OR NOT, YOU ARE A PARTY TO IT! Your paper IS being used to "create doubt" that Kerry won; and that doubt is being used to keep the information from the American public that the election MAY have been stolen. That is ALL we want! THE PEOPLE TO HAVE A CHANCE TO DECIDE FOR THEMSELVES! And HOW can they decide if the information is being kept from them! And YOUR paper (and your silence about how it is being used) IS contributing to this!

Point me to the evidence for the statement I have put in bold.

[Name removed]

170. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Febble]

182. Yes please do draw out a clear path for me

It is very hard to address the substantive comments in your posts when they are so liberally interspersed with abusive remarks regarding my motivation and integrity.

You asked two very specific questions in a thread above, and I replied.

What else do you want me to respond to? You have addressed very many posts to me, Tom, and I have mostly attempted to respond. Why I should respond to anyone who insults me in the way you do is questionable, but I will do my best. I'm taking your "Oops..." post here. If I find more I will attempt to address them too:

I think you are in "full-spin-mode" again; and, your "cognitive dissonance" protection mechanism in in full swing too.

This part is abuse.

<You said:>
""No, that is not an example of an argument" with an excluded middle, or even an assertion. It is an exhortation."

Definition:
http://www.cogsci.princeton.edu/cgi-bin/webwn2.1?o2=&o0...
"exhortation (a communication intended to urge or persuade the recipients to take some action)"

Well feeble, isn't it the purpose of ALL fallacious arguments "...to urge or persuade the recipients..."???
-----------------------------------

The excluded middle fallacious argument:

Hint: This one's REALLY EASY to spot - that is, IF one is OPEN to spotting it.

Almost any time you see an "either...or" "exhortation", you can pretty much count on there being PLENTY of excluded middle choices.

For example:

The "essence" of Bruce O'dell's "exhortation":
"...either join me...or stop questioning..."

or, GWB's:

"Either you're with us, or your with the terrorists..."

Now really, aren't there PLENTY of middles for both of the above statements? <IMO, cognitive dissonance would be the main barrier for someone to NOT recognize that there are middles in EITHER ODell's OR GWB's "argument". >

I think, since ODell is your ally in casting doubt on the exit poll discrepancy evidence, you can't be objective about this.

I assume you don't really want me to comment on this part. I will if you want.

-------------------
Ahhh, now on to your claims about what I have said:

"Bruce O'Dell does not support Baiman. Therefore he must support Mitofsky."

<So NOW we're in the realm of false logic? First, I've got to wonder (just guess) WHERE you got this idea from. So here goes:>

I said:
"That's pretty clear that he is "advocating for Mitofsky's position <re: the rBr hypothesis>"."

I don't see where Bruce is arguing for Mitofsky's position, unless you characterise Mitofsky's position as simply stating that there was no linear relationship between bias and vote-count margin. Bruce and Mitofsky probably agree on that. So do I. So, I think, does Ron Baiman. What Ron Baiman is saying now, I think, is that there may be some non-linear relationship worth investigating.

<This IS quite a leap in logic (and VERY assumptive) on your part, isn't it? How did you connect "does not support Baiman", with "must support Mitofsky"? AND, are you implying that I also said Odell "support(s) Mitofsky" overall? (That wouldn't surprise me; but I certainly wasn't even remotely Saying (or thinking about) that here.>

Fine. If you are not saying that O'Dell is advocating Mitofsky's position, or if you are merely agreeing with me that O'Dell and Mitofsky share a position regarding the lack of a linear relationship between bias and vote-count margin, fair enough. If so I don't understand your problem with O'Dell's position.

"Febble does not support USCV. Therefore she must support rBr."
<WRONG AGAIN. I derive your support of rBr from your advocacy of Mitofsky in using "your" function/paper to support rBr. <BTW: Just because you have sent PLENTY of "mixed messages: regarding this, doesn't really change that fact, does it?>

I have not advocated Mitofsky in using "my" function. I suggested, publicly, in my paper, that E-M reanalyse their data using my function. Mitofsky did so. The analysis does not prove rBr. It only supports it in that it refutes the claim that "Bush strongholds have more vote-count corruption". Or at any rate, O'Dell, Mitofsky, and I appear to share the view that it refutest that claim. It does not mean there are not other plausible fraud hypotheses, as O'Dell suggests.

And, finally, you said:
"...a major problem on this forum...intolerance of any analysis...shows any sign of conflicting with the case that Kerry won the election..."

That's your ACTUAL "bottom line", isn't it REALLY?

And YOU should be able to understand WHY "we" have such a problem with this: This has been the modus operandi of the attempted cover-up since the election was stolen, hasn't it? Create doubt that KERRY REALLY WON, wherever, and whenever possible.

I've already answered this. Yes, I doubt that Kerry won the popular vote. I am not creating doubt, though I am sharing my grounds for that doubt. But I also believe, which is why I wrote my paper, that until the data is reanalysed using something like my function (something better than the WPE anyway) that the matter will remain in doubt, either way. And the fact that it is in doubt at all - that it is possible to believe that Kerry won the popular vote, is itself a scandal. We can, mostly, trust our voting system in the UK. You can't. That needs fixing, whowever "really" won the 2004 election.

And, creating doubt that the exit poll discrepancy evidence points to a stolen election has been in full gear ever since January, when Mitofsky re-fabricated rBr for the son (GWB - like he did for the father in '92). Pastor used this rBr as an excuse to keep USCV, and the exit poll discrepancy evidence, from being heard at the first Carter/Baker Election Reform Hearing. And, is there ANY doubt that your paper is being used to further justify USCV NOT getting to testify at the second hearing (coming up on June 30th)?

.

I simply do not believe that Mitofsky fabricated anything. I think you are simply wrong on this, and that your assertion is not supported by the evidence. As for my paper being used to justify USCV not getting to testify, please cite the evidence. Bruce O'Dell may have some comment to make as well, as VP and Press contact for USCV.

I've suggested this to you before: I think you really need to do a "reality check". Considering how you responded to my "excluded middle" question, AND what I just reminded you about, it's time to bypass your ego (and wishful thinking), before it's too late - TIME IS RUNNING OUT!

My response to your excluded middle comment was made in anger- you have repeatedly on this forum twisted something I have said to imply support for a position I do not embrace. As someone who inhabits middle positions on many things, I have strong views on arguments with excluded middles.

If I misrepresented your position, I apologise. You have certainly repeatedly misrepresented mine, and your incessant attacks on my personality, motivation, ego, and integrity are unfounded and unjustified.

You and I disagree on our interpretation of the evidence. Let's leave it at that.

[tommcintyre]

220. My response

Febble, this is the longest post I have ever written. I’m investing this time in you, hoping to “get through”.

I do believe the paper you wrote IS being used to do substantial and serious damage to the chances of even getting the possibility that the election was stolen before the US public. And, this may also consequently effect our chances of achieving serious election reform.

I also believe that your acquiescence to allow your paper to be used in such a way, puts you in a position of aiding and abetting this theft and cover-up.

As promised, you will find my "case" for why this is true at the end of this post. When finished, I will post "the case" on Monday (with some additional appropriate information) as a separate thread since I believe it is so important to our ongoing efforts in this forum.

Also, I will PM this entire message to you, since the thread it is going in is becoming very ungainly.

<My comments in <...>

<One suggestion, view my comments more as feedback, rather than critisim. My replies are what I honestly thought when I read each line. If you don't take it too personally, it's a rare opportunity to get unvarnished feedback. Maybe, imagining it's written about someone else rather than you may help too, to increase objectivity. Also, of course, I'm not always right. ;) >
---------------------------------------------------------------------

It is very hard to address the substantive comments in your posts when they are so liberally interspersed with abusive remarks regarding my motivation and integrity.
<Very vague, broad, unsubstantiated innuendo - misleading.>

You asked two very specific questions in a thread above, and I replied.
<Avoiding the issues on a technicality (address questions only), which you demonstrate you are fully aware of, since you "get down to business" later (at least partially) and address the substance of the post.>

What else do you want me to respond to? You have addressed very many posts to me, Tom, and I have mostly attempted to respond.
<Do you respond openly, honestly, fully - or do you "cherry pick" those points you personally feel more comfortable with, and ignore/avoid those which you do not?>

Why I should respond to anyone who insults me in the way you do is questionable, but I will do my best.
<Again, vague unsubstantiated claims of being "the victim", followed by a "demonstration" how amazingly "fair and tolerant" you are.>

I'm taking your "Oops..." post here. If I find more I will attempt to address them too:

I think you are in "full-spin-mode" again; and, your "cognitive dissonance" protection mechanism in full swing too.

This part is abuse.
<WOW! Do you really think the "average person on the street" would find THIS abusive? This is merely a very accurate description of what I believe you are doing. If I said, "you're an idiot (etc.)..." - THAT would be abuse. Yes, I know abuse can be more subtle than that, but the "idiot" test is a pretty good one. Am I commenting on you directly; or using clearly abusive labels (idiot, etc.)? A simple example (to a child): "YOU are bad!" (abuse); or, "what you did is bad!" (no abuse).

So, you interpret criticism as abuse. I've seen you do this on more than one occasion - with different people. For example, if someone expresses the opinion, "you are spinning", you feel abused, don't you? Isn't this REALLY just criticism? (I'm not saying you have not been subject to abuse at different times. If you post enough here; and/or take "controversial" positions - it's bound to happen.)

Hint take ALL criticism as feedback, and detach your ego from it - this is a VERY USEFUL thing to do.>

I assume you don't really want me to comment on this part. I will if you want.
<Re: The excluded middle section of my post:>
<I let out a big sigh of exasperation when I saw you skipped this part - I'm glad to see you thought the better of it, and at least partially addressed it later. I will comment more on it "as it comes".>

I don't see where Bruce is arguing for Mitofsky's position...

<from Bruce's paper:>
"...when the Liddle Bias Index is applied to the “USCV O’Dell simulation” data cited in the Working Paper, it produces results consistent with those recently reported by Warren Mitofsky for the E/M data as a whole."
<This could easily be a miscommunication/misunderstanding (communication breakdown). Aggravating the situation is 1) Bruces' strongly adversary tenor ("fatally flawed", etc.), assuming knowledge of his paper (implied), of the technical terms, etc. Considering all this, I took the most common meaning for "consistent with" (in agreement with) as his broad intent regarding opposition to the USCV paper; and aligning himself with the countervailing Mitofsky position.

I was frankly surprised that you (and Bruce) interpreted this to mean that I think he supports Mitofsky "carte blanche" (his methods, questionable history, etc). At this point, I couldn't possibly know that - but I wouldn't be surprised if he leans more toward him, than away, as you and OTOH appear to be (that would be "consistent", wouldn't it? ;) ).>

I have not advocated Mitofsky in using "my" function. I suggested, publicly, in my paper, that E-M reanalyze their data using my function. Mitofsky did so. The analysis does not prove rBr.
<From a post by RonB:
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x372464#372912
"This is important as E-M <Mitofsky> presented your work as a breakthrough that proves the rBr hypothesis..."
THIS IS HUGELY PROBLEMATIC! Are you willing, RIGHT NOW, to CLEARLY come out against this position, and fight PUBLICLY to stop your position from being used this way? And please, don't quibble this point: I know it, Ron knows it, AND YOU KNOW IT IS TRUE.
(Are you beginning to get the connection of how your function/paper is being used by Bush/Mitofsky et al to keep USCV from testifying at the hearing?) BTW: You "advocate" (if by nothing else) by your silence (implied agreement) on this matter (the complicity of silence - for example, many Germans did that in WWII when the Jews were taken away, didn't they?) Yes, Lizzie, you ARE responsible.>

Yes, I doubt that Kerry won the popular vote. I am not creating doubt...
<It really doesn't matter what YOU believe - that's NOT the issue. Have I ever said I was upset with you because YOU think this? NEVER! And, I would even defend your right to believe this way (the Voltaire/Patrick Henry thing). I'm serious. If anyone EVER tells you here that you can't personally believe this, I will come and explain to them why you have this right.

I'm NOT in the conversion business, I'm in the EXPOSURE (of the lection fraud information/evidence) business. I've said this to you before, and (apparently) it just doesn't sink in. In my "Of course!" post, here is what I just told you:
"That is ALL we want! THE PEOPLE TO HAVE A CHANCE TO DECIDE FOR THEMSELVES! And HOW can they decide if the information is being kept from them!"
Can I be ANY clearer than that? (You should ask yourself, WHY this msg didn't "take" with you - could it be cognitive dissonance? This should be more "alarm bells" for you - is it?)

Also, again, the REAL problem is NOT what YOU believe; but how your paper is being used as part of the effort "to keep the information from the American public that the election MAY have been stolen." (Again, a quote from my "Of course!" post.)

So, again your acquiescence by silence is the real problem".
I said, in the same post:
"And YOUR paper (and your silence about how it is being used) IS contributing to this!"
snip
"YOU SHOULD BE DOING EVERYTHING YOU CAN TO REFUTE HOW MITOFSKY IS USING YOUR PAPER; AND HELP TO GET USCV A CHANCE TO TESTIFY AT THE UPCOMING CARTER/BAKER COMMISSION ON JUNE 30TH!"
Is there ANYTHING that is REMOTELY ambiguous about THESE statements?
(If not, again, WHY aren't you receiving "the message"? Again, could cognitive dissonance have ANYTHING to do with it?)>

I simply do not believe that Mitofsky fabricated anything. I think you are simply wrong on this, and that your assertion is not supported by the evidence.
<I don't know whether he fabricated anything either (not enough information to judge - but the reweighting IS a little "convenient" though). I never said he "fabricated" the data this time. What I did say was:
"2- And even your inferred data that you have to work with is so fatally incomplete (weak); since the NEP and/or Mitofsky won't release the "raw data" (and even if they did, HOW would we ever know it wasn't cooked?)."
So the real question is, HOW did YOU construe that I was sure Mitofsky fabricated the data from THIS? (Can't you now see why I CLEARLY think you have a bad case of cognitive dissonance? That one, you can take to the bank! BTW: The only chance you have of getting a handle on it is to admit you have it every day - and mean it! That's the conclusion I came to after literally years of thought on the issue. I saw from a very young age (about 6) what a terrible problem what is commonly known as denial, or rationalization really is. This is why the Bob Woodward quote literally jumped out of my head when you asked if I ever doubt. I have made a life-long study of all aspects of the phenomenon - from the bad to the good.

Example of the bad: Facilitated Communication (for treating Autism - something you may find interesting - link below - thank who/whatever for the scientific method/double-blind study).

The good: terminal "hopeless" cancer patients: quite a few years back, when I wrote some books on "optimizing health, nutrition and longevity", I came upon this study: survival rates of breast cancer patients were surveyed: 70% survival for the "proactive" patients - 20% for those who "gave up the ghost". No surprises here - the purpose of the article (where I got the quote from) was to point out how important a positive mental attitude is in good health. But, here's the kicker: 50% survived that went into total denial! (Over twice as many as the "resignationists"; and only 20% less than the best group.) Also, I think animals use "denial" (playing possum) as a last-ditch effort at survival when all other hope appears to be lost. (Another study (apropos to this forum's purpose) revealed that the illusion of freedom (democracy?) was almost as effective as having actual freedom. This could help to explain a lot of so many US citizens’ attitudes/inaction. ;) )

So, the conclusion: denial (cognitive dissonance) is a useful tool as a LAST RESORT; but unfortunately is all-to-often used when other proactive options are available.

The autism-related link:
http://www.pbs.org/wgbh/pages/frontline/programs/transcripts/1202.html
Prisoners of Silence (If at all possible, I recommend you see the video. It is truly remarkable to see the reactions of the "professionals" (who had so much "invested" in this - especially emotionally) when a double-blind test debunked the entire process. This is truly a revealing lesson in just how serious, pervasive, and susceptible we really are to cognitive dissonance.
----------------

Regarding Mitofsky:
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x371726#373177
(I've pasted the most important parts here:)
"...you revealed some interesting things about Mitofsky:

1) He claimed rBr as fact in his original report. At MOST, this was a hypothesis, he knew that, EVEN his rBr "supporters" call it that. Plain and simple, legitimate professionals NEVER claim a hypothesis as a fact - by definition, it is simply a "guess", or "hunch" that must be investigated further. THIS ACTION ALONE IS A STRONG BREACH OF ETHICS. Thanks for pointing it out.

2) As you also pointed out, he also advanced rBr WITHOUT ANY SUPPORTING EVIDENCE. Again, this is something no ethical professional would dream of doing, in a formal report, as he did. They would wait until the investigation was reasonably done, and ONLY THEN publish their findings. Again, Thanks for pointing it out.

---------------------------
Febble, would you at least listen to this interview? I've read your paper twice, printed it out, and highlighted it and notated it, so this isn't too much to ask, is it? ;)
3. I've listened to Victoria Collier's interview, and highly recommend it

http://www.edwardsdavid.com/BushVideos/infidelguy.com_Vote_Scam_a.mp3
Also, read the first eight chapters (three more than at the votescam site) of the book here:
http://www.constitution.org/vote/

*<Verification here:>
http://www.votefraud.org/Archive/Write/newhampad.htm
"...early media reports stated that Bush and Buchanan were running neck and neck in the 1992 New Hampshire primary..."

<My original source:>
http://www.votefraud.org/News/2000/7/071800.html
-----------------------------------------------

And this is just the tip of the iceberg. Mitofsky has been at this since the mid-sixties.

I've found a clear pattern of <from another post of mine:>

http://www.democraticunderground.com/discuss/duboard.ph...
"Mitofsky uses a number of methods to "cover-up". It seems to range from computer meltdowns, "found votes", withholding the data, "convenient guesses", etc.

So what may have happened in 2004? <To "trip up" Mitofsky.>

- More scrutiny because of publicity of the 2000 theft.

- Internet leaks"

<Some other related links:>

Election Night Projections - Cover For Vote Rigging Since 1964?
http://www.ecotalk.org/VoterNewsService.htm

2002 Elections: Republican Voting Machines, Election Irregularities, and "Way-Off" Polling Results
http://www.ecotalk.org/MidtermElections2002.htm

Suspicion Surrounds Voter News Service
http://www.ecotalk.org/VNSclosingdoors.htm

Could the Associated Press (AP) Rig the Election?
http://www.ecotalk.org/AP.htm

Votes Downloaded to AP - gets direct feed from voting tabulating computers
http://www.ecotalk.org/APvotes.htm

<You know, it was YOUR assertions of how "squeaky clean" Mitofsky is (in Kathy Dopp's thread the Thursday before Mitofsky's presentation) that tipped me off to check him out. (I was getting really "bad vibes" by what you were doing in that thread, so I thought I better check him out.) I really didn't have much of an opinion on him, either way, before that (just a mental note to eventually check him out, because of his appearance in the Jan 19th Nightline show). Once I started checking him out that Thursday, the "dirt" just started pouring in!

I'm curious, do all three of you (OTOH and Bruce) generally hold Mitofsky in such high regard? If so, is there a pattern here?>

My response to your excluded middle comment was made in anger...
<Uh oh! So what does that tell you? If you have an "emotional attachment (involvement)" in an issue, are you really able to make objective choices? (i.e. without cognitive dissonance "clouding your vision".) Can YOU detach yourself enough to see the pattern I'm seeing (and have been seeing) with you?

In contrast, I'm VERY angry with your actions, and the potential consequences of them. (not, with you - but with your actions). I can separate who you are, and what you do. So, as I mentioned above, if someone says you don't have the right to believe Kerry didn't win the popular vote, I'll defend your right to believe that (but NOT your opinion, of course ;) ).

So, what's the difference in how we are processing the two situations? I'm detaching myself emotionally from the situation (being objective - avoiding cognitive dissonance) - you, are not. What I have said is NOT to "get you", "attack" or "abuse" you. It is simply in hopes of getting you to increase your objectivity in what you do and say, regarding election fraud and reform.>

BTW: "...you have repeatedly on this forum twisted..."
<Didn't you just blame me for what you did (at least sort of? ;)>

If I misrepresented your position, I apologise. You have certainly repeatedly misrepresented mine...
<Didn't you just "apologize", and then "take it away? (an "un-apology"?) I think apologies are really secondary (I really wasn't looking for one). What's important is the right actions are realized - and an apology can be a "sign" of that - but, clearly, not always.>

You and I disagree on our interpretation of the evidence. Let's leave it at that.
<I truly wish that's ALL there is to it (like disagreeing on whether Kerry won the popular vote); but CLEARLY there IS much more to it than that, isn't there? If your paper IS being used to suppress the possibility that Bush actually stole the election - there IS much more to it than that, isn't there?

=====================================================================

How your paper is most likely being used to suppress USCV's exit poll discrepancy evidence, and how it is being used to keep the US public from knowing about the possibility the election was stolen:

1- The Carter/Baker Election Reform Commission had it's first hearing (April 19th). There was NO mention (or testifying panelist) on the exit poll discrepancy evidence (which indicates that there may have been significant fraud - enough to have cheated Kerry out of the presidency).

2- The day after the hearing, Robert Pastor (the director of the commission was interviewed on C-Span. There was a portion of the interview where he took caller's questions. When a caller asked him why Steve Freeman (and the USCV evidence) was NOT included in the hearing to present the exit poll discrepancy evidence, Pastor quickly dismissed the caller by saying that the evidence had been discredited. (This is mainly what caused the thousands of angry emails that Pastor explained he received, when he called BradBlog to complain he was being harassed by these emails.)

3- Meanwhile, USCV had come out with an even stronger paper, indicating the possibility of election fraud.

4- Shortly thereafter. Elizabeth Liddle - a psychology PhD candidate(known as "Febble" here on DU), came out with her own paper which was used by Warren Mitofsky (the head of the group that did the exit polls) at a statistical conference in mid-May. Mitofsky used the paper to further support (prove) that the exit poll evidence did NOT point to election fraud.

Conclusion: Since the original claim that Mitofsky made that there was no significant election fraud (with no supporting evidence, when he released his original report on January 19th) was primarily used to justify NOT including the USCV (Freeman) exit poll discrepancy evidence; it is reasonable to assume that Febble's paper will be further used to block the exit poll discrepancy evidence from being presented at the upcoming hearing on June 30th (the last one).

Of course, we shouldn't be expecting anyone to come out from the C/B Commission and admit to this, should we? ;)

[Febble]

227. Copy of PM to Tom

Sorry, Tom I did not realise you had posted this. Here is my PM response, and some additions.

First of all, Tom, thankyou for sending the PM.

We clearly see most things very differently, but I appreciate that you are honestly trying to "get through". I think we are both honest people.

Here is a response, which you will find inadequate, I'm afraid. The gulf between us simply looks unbridgeable. I'll take a few points:

1. Regarding criticism: I am distinguishing between what I am prepared to call critique and abuse. I am happy to respond to argument. Where I find your comments simply abusive is when you imply that I am being dishonest, afraid, motivated by ego, or whatever. That is what is called "ad hominem" argument, as you know, and it is quite different from reasoned debate. I realise you think your remarks in this vein are "constructive criticism". However, they are, nonetheless, criticism of me, not of my arguments, and thus ad hominem. In an online public debate, where neither of us no each other, assumptions as to other people's motivations, it seems to me, should be made with caution. It may not look like abuse to "the man on the street". But ad hominem debate is a form of abuse.

2. If my paper is being used in the way you claim, then it is being used wrongly. All arguments should be made on their merits. I think my argument stands on its merits. If it is defeated by a better argument, fair enough. As you have read my paper, you will know that it is primarily levelled at the measure of bias used by E-M in their report. And it shows that because of the flaws inherent in that measure of bias, the inferences made in the original USCV paper from those E-M measures cannot be sustained. I realise that Ron Baiman still thinks the inferences are correct, and he is of course entitled to make his argument. This is not a matter of politics. It is a matter of numbers. I still think Baiman's interpretation is incorrect. It does not directly support "rBr". It is simply "consistent with it". It is also consistent with fraud that is not concentrated in Bush strongholds (and actually, I now believe as a result of online discussion on DU, more consistent with fraud concentrated at the Kerry end of the spectrum).

3. If something is in the public domain, I am in no position to "acquiesce" or otherwise to the uses to which it is put. If it is being used to bolster a case, and it does not bolster that case, then someone needs to show that it does not. It does bolster the case that fraud was not more prevalent in "Bush strongholds" (see 2 above). Saying this does not aid and abet anything.

4. You asked me to respond to a post, but I was not at all sure which you meant - I did my best not to cherry pick. If I missed something, please PM me. You may not like my responses, but I was not attempting to avoid any questions.

5. Regarding the release of the data. I do NOT support the release of the data. I would certainly support any application for release of "scrubbed" data, for independent analysis, as has been done for Ohio.

As for the rest, I have read the rest of your PM, Tom. I have attempted to address the points you made before. You are angry with me, and there is little I can do about that. There is simply no answer I can give that will satisfy you. I know this is frustrating for you.

I think there is only one answer, and that is for me to stop debating on DU. I will continue to respond to any PMs. But it is time I stopped expressing views on that forum. I came to debate my paper, and stayed too long. I have had some good debates, and learned a lot, much of which has been useful to the development of my thoughts regarding the way the data should be analysed, and the kinds of patterns that might indicate fraud.

So in that sense, I am glad I stuck around.

I'll read your piece when you post it.

And I wish you, truly and honestly, the best of luck with your campaign to reform the US electoral system. It certainly needs it.

Lizzie

I will say publicly that my work in itself does not prove "rBr". It enabled an analysis to be done that refuted a specific fraud hypothesis, made by USCV in their report: "Bsmvcc" = "Bush strongholds have more vote count corruption". I do not believe that differential response bias can ever be "proven" by inferential statistics, any more than fraud can, and certainly not by a single correlation. But I do not believe that what we know now about the data is "inconsistent" with differential response bias. I therefore disagree with the Baiman Dopp paper that other explanations "are required" to explain the data; this is not to say that other explanations may not be valid.


I will also repeat here that I think it is better if I stop posting. I will do a short piece explaining the reasons, and will welcome any PMs that anyone wants to send. I will try to respond. But I am aware that my presence here is simply divisive.

[autorank]

230. Before you stop posting, why don't you finish up on the TIA rBr thread

That's a priceless opportunity toi walk the walk and engage in the type of dialog you say you want. Failing to finish there would leave some with the wrong impression. Then you can do you PM thing.

[Febble]

183. OK, I think I've tracked all your comments

Re excluded middles you wrote:

<You said:>
""No, that is not an example of an argument" with an excluded middle, or even an assertion. It is an exhortation."

Definition:
http://www.cogsci.princeton.edu/cgi-bin/webwn2.1?o2=&o0...
"exhortation (a communication intended to urge or persuade the recipients to take some action)"

Well feeble, isn't it the purpose of ALL fallacious arguments "...to urge or persuade the recipients..."???
-----------------------------------
The excluded middle fallacious argument:

Hint: This one's REALLY EASY to spot - that is, IF one is OPEN to spotting it.

Almost any time you see an "either...or" "exhortation", you can pretty much count on there being PLENTY of excluded middle choices.

For example:

The "essence" of Bruce O'dell's "exhortation":
"...either join me...or stop questioning..."

Arguments, whether fallacious or not, are propositions. If they have excluded middles they are fallacious. Bruce's exhortation was not an argument. It may nonetheless have had an excluded middle. Many exhortations do. It's supposed to put you between a rock and a hard place.

or, GWB's:

"Either you're with us, or your with the terrorists..."

Now really, aren't there PLENTY of middles for both of the above statements? <IMO, cognitive dissonance would be the main barrier for someone to NOT recognize that there are middles in EITHER ODell's OR GWB's "argument". >

This is an assertion, not an exhortation. It has an excluded middle and is fallacious. Phrased as an exhortation it would be "either join us, or join the terrorists". It would be unreasonable, but not illogical. As an assertion it is illogical. It is perfectly possible to condemn both the terrorists and condemn Bush. I do both, as I am sure you do too.

I think, since ODell is your ally in casting doubt on the exit poll discrepancy evidence, you can't be objective about this.

Possibly not. But he became an ally because we came to a similar conclusion. We did not come to similar conclusions because we were allies.

After all, he was an author of the paper my paper critiqued. He is also VP of the organisation that produced a paper billed, here on DU as a "rebuttal" to mine, as well as creator of the simulation that was going to do the "rebutting".

[Name removed]

148. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

149. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

151. Deleted message

Message removed by moderator. Click here to review the message board rules.

[LightningFlash]

153. Actually Febble, I believe you need to practice what you preach.

I have always been one of the largest doubters that Kerry really won the election, until I saw the full weight of the evidence, that is and became disassociated from that rhetoric.

http://www.votersunite.org

In all instances, more than 80% of the touchscreens switched their votes, Kerry to Bush. Accumulation of them coming from Ohio, even though Florida was the first to report of it.

Then of course, Febble, you must further put doubt in your own analysis. You aren't acting like you are interested in the truth, when you ignore much needed and uncontestable factors such as this.

http://www.census.gov/Press-Release/www/releases/archives/voting/004986.html

The bogus party weighting, and in-consistent number reversal could not be due to poor sampling bias which is called "random sampling error" Febble, and it could never be explained as such.

In addition to those common facts, the US Government's released data shows an archetype of how many voted and who voted nearly everywhere. Your measurement tool, nor the ones currently supplied, offer any kind of a conclusive way to actually "show" how the fraud ocurred, and that is more in due to poor testing on their part than anything else.

That does not mean under any circumstances, that it has shown fraud didn't occur. In fact, if the vehicle is as hastily compiled as this one was no one would even be able to reach such a conclusion and certainly no "non-citizens" of the U.S.

The official measuring tool for the 2004 election, which is more reliable and indeed stable then that of any of the so called measuring tools to date is the U.S. Government's own cencus bar.

Why don't you go start an academic debate with the government instead of wasting your time trying to save Mitofsky's face for the people? Mitofsky's already finished, as per due to his "poor, garbage" exit polling - and poor misguided measurement tools. Soon exit polls will be outlawed anyway Febble, and you have done absolutely nothing but receive some money from Mitofsky's polling firm Voters News Service.

So how do you feel, exactly? The barometric pressure has now become unbearable. In fact, election fraud did occur. It truly did tilt the election, but it can not be shown by how much or how little until ALL the precincts are re-weighted.

Where's the fancy function for that I wonder? Also, whether Kerry won by directly one million votes or 6 million votes does not change the facts- There was clear cut and dry exit poll results dropped from the final re-weighting period. Exit poll results in twine, that favored the challenger but did not get reported. And that will remain not only the bane, but the crime which isn't accounted for.

[Internut]

154. Can you point out what exactly from

the census.gov press release convinces you that the exit polls were tampered with.

[LightningFlash]

161. Table 4-A, Exhibit of National Election 2004 United States.

Need more collaboration? Examine http://exitpollz.org directly alongside of the Cencus' results in a cross section.

[Internut]

187. The link you gave -

http://www.census.gov/Press-Release/www/releases/archives/voting/004986.html

does not contain table "4-A". Please point to the specific table you meant.

[LightningFlash]

201. Can be studied here, and here.

Can be studied here-
http://www.census.gov/population/socdemo/voting/cps2004/tab04a.xls

As well as the rest of the reported voting registrations by Age, Sex, and Catergory as outlined here.

http://www.census.gov/population/www/socdemo/voting/cps2004.html

[Internut]

205. Ok, let's study that link:

Demographics:

Census Survey Data          | Final Exit Poll 
(from the "Voted" column    | (13,660 respondents) 
----------------------------|-------------------------------------
Male    58.45/125.73  = 46% |     46%  
Female	67.281/125.73 = 54% |     54% 
                            |
White   106.59/125.73 = 79% |     77% (36+41)
Non-White               21% |     22% (10+12)

That's a pretty good match. What exactly are you seeing in the census data that would show that the final exit poll's numbers are "cooked"?

[LightningFlash]

209. And what of the final data?

How was it's weightings not cooked, may I ask?

http://www.exitpollz.org/CNN_national1253.html

" 1. VOTE BY GENDER

BUSH

KERRY

NADER

TOTAL
2004
2000
2004
2004



Male (46%)
55%
+2
44%
0%



Female (54%)
48%
+5
51%
0%






2. VOTE BY RACE AND GENDER

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



White Men (36%)
62%
n/a
37%
0%



White Women (41%)
55%
n/a
44%
0%



Non-White Men (10%)
31%
n/a
67%
1%



Non-White Women (12%)
24%
n/a
75%
0%






3. VOTE BY RACE

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



White (77%)
58%
+4
41%
0%



African-American (11%)
11%
+2
88%
0%



Latino (8%)
44%
+9
54%
1%



Asian (2%)
42%
+1
58%
*



Other (1%)
41%
+2
54%
2%


13. VOTE BY EDUCATION

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



No High School (4%)
50%
+11
50%
0%



H.S. Graduate (22%)
52%
+3
47%
0%



Some College (31%)
54%
+3
46%
0%



College Graduate (26%)
52%
+1
46%
1%



Postgrad Study (17%)
45%
+1
54%
1%


15. VOTE BY PARTY ID

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



Democrat (37%)
11%
+0
89%
0%



Republican (37%)
93%
+2
6%
0%



Independent (26%)
48%
+1
50%
1%


17. HAVE YOU EVER VOTED BEFORE?

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



No (11%)
46%
+3
53%
0%



Yes (89%)
51%
+3
48%
0%






18. VOTE BY RELIGION

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



Protestant (54%)
59%
+3
40%
0%



Catholic (27%)
52%
+5
47%
0%



Jewish (3%)
24%
+5
75%
*



Other (7%)
24%
-4
74%
2%



None (10%)
31%
+1
68%
1% "

They seem to be polar opposites, a switcharoo of that magnitude, combined with affiliations shows a bogus re-weighted bias. A bias towards male bush voters, and you are not stating your case very well are you.

[Name removed]

210. Deleted message

Message removed by moderator. Click here to review the message board rules.

[autorank]

184. febble, Tom's point on doubt CRIES OUT for an answer, not just from

you but from OTOH, O'Dell, and Time for change. At this point, the doubters by their relentless self-reinforcing attacks on the validity of election fraud (through narrow hypotheticals like rBr) are enabling Bush and co and, most certainly, the person who helped Bush steal Florida in 2000, Baker.

WE'LL NEVER GIVE UP BUT TOM'S RIGHT, TIME FOR YOU FOLKS TO GET ON BOARD ONE WAY OR THE OTHER.

As of right now, you are the great enablers of the election fraud debunker movement...a vital piece of maintaining Bushco's power.

NEW LEADERS FOR A NEW DEMOCRATIC PARTY!

Contact the DNC and Give 'em Hell About NOT Acting on Election Fraud

[Febble]

190. If I understand you and Tom

You don't like this debate because it "creates doubt".

I was not happy with the term "create" because it implies that I, or any of the others you cite, are deliberately and maliciously sowing seeds of doubt as opposed to...well, simply doubting.

It seems to me there are three campaigns being promoted on this forum:

First: to reform the voting system so that in future no-one will have any doubts as to who won the election. With that I completely agree, and have done what I could to promote that cause. It is at the core of USCV's mission, and is why I worked with them for so long. No doubts there.

Second: to demonstrate that Kerry won the election. This hinges largely on Ohio. Exit polls simply will not give us the answer to this - there were far too few precincts sampled in each state. This needs to be done on the ground. I have already contributed a bit to the analysis of voting machine provision, and believe that through either incompetence of malice (from a post above it looks more like incompetence) far more Kerry voters were disenfranchised in Columbus than Bush voters. Also there was no re-count, and Blackwell's behaviour is suspicious at every point, and prima facie illegal at some. But this an argument for an investigation (and possibly prosecution). I also did a bit of an investigation into undervotes in New Mexico, which may have cost Kerry a few more electoral college votes (see USCV website). I completely support the campaign for investigation into electoral fraud, and voter suppression. I don't think it is necessary to believe Kerry actually won Ohio to support this campaign - and if voter suppression was key, then he simply may not have had the votes, even if he should have had the votes. Provisional ballots were another scandal.

Third: to demonstrate that Kerry won the popular vote. Again, this should simply be an investigation. No investigation should be hampered by a priori assumptions as to the answer. If it is simply a political cause, to be advanced regardless of evidence to the contrary, then the train leaves without me.

I have never been happy with faith-based politics, even in the days when I marched with the hard left in London for the end of apartheid, or unilateral nuclear disarmament, and had to opt out of some of the more extreme chants ("No free speech for Fascists!"). I'd love to believe that Kerry won election, as you know. It's why I'm here. But I don't think the exit polls prove it (or disprove it, but certainly don't prove it). And I'm not going to say they do if I'm not convinced they do. I think TIA is mistaken.

And I simply do not believe that the case for fair, secure, auditable elections hinges on whether Kerry would have won or not. It is simply fundamental to a democracy.

To that extent, I agree with you that this exit poll discussion is irrelevant.

[Time for change]

195. I have one slight disagreement with what you said

Whereas I agree that the case for fair, secure, auditable elections SHOULD NOT hinge on whether Kerry won the election, in practice it may well hinge on that. That is because the American public is so complacent about this that it may be that the only way to make them wake up to the need for election reform is to be able to demonstrate that the election was stolen. I think that this opinion is born out by the way that Congress is dragging its feet on meaningful election reform. That is why I think this whole issue of analyzing the exit polls is so important.

[autorank]

199. I love debate.

I was a successful high school and college debate team member, really rocked. No problem there. What I have a problem with is ... well, it's the deliberate waste of time created here. Instructive, needs to be challenged but a waste of time.

1) You say you are for election reform and then: "It is at the core of USCV's mission, and is why I worked with them for so long. No doubts there." Well, you, O'Dell, Time for change (yes you Tfc, the secret is out) and OTOH certainly trashed and embarrassed USCV, right before the Carter-Baker hearings. How convenient.

2) You say "Second: to demonstrate that Kerry won the election. This hinges largely on Ohio." You're completely wrong! How about Florida, New Mexico, and * inflation in heavy Kerry states like CA and NY. There is information here, in detail, on this. If it were just Ohio, it would be an easier process. I'm truly surprised you would make such a clearly inaccurate statement.

3) You say that to demonstrate Kerry won "this should simply be an investigation." Wow, "simply be an investigation." And how might that happen, with you, Time for change, OTOH, and O'Dell here trashing much of the basis of an investigation.

Your first argument is contradicted by your attacks on the organization (USCV), your second clearly absent any connection to the reality of election fraud accusations, and your third argument is naive.

Debate indeed. You only want to debate when you outline the proposition solely on your terms and set the rules in the style of the Queen of Hearts.

[Febble]

200. OK, have it your way.

I'll stop debating.

[kiwi_expat]

206. If Kerry won Ohio, Kerry won the election.

"* inflation in heavy Kerry states like CA and NY" is only relevant to demonstrating that Kerry won the POPULAR vote.

You say "If it were just Ohio, it would be an easier process". Are you referring to recounts, or do you have some other analysis in mind. All suggestions as to how to demonstrate that Kerry won Ohio will be gratefully received.


My biggest concern about all of the attention given to the POPULAR vote is that it implies that the winner of the popular vote WON the election.

Bush "won" the election by little more than 100,000 votes. And there is a good chance that we could PROVE that he did not get those 100,000 votes. Let's focus on that!!

[autorank]

207. You make a very good point. I was getting at something a little

different but I had the implication there. Probably because it would have been really problematic if Kerry had received less popular votes but won Ohio. What I meant to say is that if Ohio were the only ground zero in the election fraud movement, there would be, at least some boundaries. I don't need to go into it, but there are many other places with problems, hence the great difficulty. I have faith in the people working on this in Ohio and the others helping that the truth will out. I have more faith that if the Democrats take Attorney General (elected in OH?) and, more certainly, the governor's mansion, then we will know the truth. Nothing like a REAL INVESTIGATION with sworn testimony to get people to tell the truth. With Taft at 17%, maybe we have a shot.

[Internut]

208. New Mexico - the Governor, the

Attorney General and the Secretary of State are all Democrats. Did that help?

[helderheid]

229. The voting machines

If those states used Diebold or ES&S or any other machine whose companies support * and the GOP, then yes, they did "help".

DOH!

I promised I'd leave this freakin' distractor thread alone.

BAD HELDERHEID, BAD!

[OnTheOtherHand]

231. ah, that answers my question on the other thread

Look, it is not my fault, or Febble's, or O'Dell's, that the president of USCV insisted (pretty much at the last minute) on releasing a working paper with what all three of us believe to be gaping holes. It is pretty wishful to imagine that if the three of us put a happy face on the problems, no one in the real world will notice them.

Dare I ask what secret about Tfc is out, with respect to trashing and embarrassing USCV? I think I missed that.

[tommcintyre]

53. I'm more interested in your posts. After all, that IS what we do here

Eventually, I will read your paper (time is tight right now - I still have to finish my critique of our original post - tomorrow); but this forum (like most I imagine) is about reading each others posts, and replying to same.

So you see, if you say one thing here; but really mean for us to understand what you wrote somewhere else, you won't be communicating very effectively. Providing links to support what you've said here is a great way to support your statements; but you still really need to provide us with enough information in your posts to evaluate your arguments. And, if what you write here conflicts with (or contradicts), what you have linked to, you have only yourself to blame for any "misunderstanding" that may ensue.

And, BTW, telling us <from your original post>:
"I've written a paper that addressed this issues in detail, that can be found at..."
is NOT clearly telling people they need to read your paper to reasonably understand your POV, is it?

And,considering all of this I still stand by my responses to what you actually communicated in your post. If you have issues with specific points I made, be specific, instead of using broad innuendo.

Now, on to the remainder of this current post of yours <your statements are in quotes, mine will be between <...> >:

"Actually, your postings about my paper draw liberally and adundantly from the "List of Fallacious Arguments" you so helpfully provided."
<I'm glad to see you actually read it; although, from your statements below, it looks like you missed the purpose/spirit of it all (more on this later).>

"-Ad Hominem: you can't refute what I say."
<I did, very clearly.>

"Therefore you attack me as a disruptive influence on a group of people you appear to admire."
<You were "attacked"? Not really - I criticized your statements - you need to point out specifically where, in case there is any misunderstanding.>


"-Argument by Laziness: if you had read my paper, you would not say that I "supported Mitofsky" (whatever that means)."
<As I said, I based my response on what you wrote in your original post. I really can't help it if you contradict yourself somewhere else, can I?>

"-Appeal to Widespread Belief: since you and many of your associates apparently already know that vote fraud occurred, you don't need "proof"."
<WOW! THAT is a remarkably assumptive (and unfounded) statement.>

"-Argument by Emotive Language. Ah, far, far too many to choose from."
<You really DID miss the spirit of the fallacious arguments list, didn't you? Making unsupported biased claims like this is a REAL NO-NO.

<But here, I'll give you an example of YOUR emotive (manipulative?) language (from another post in this thread):>
"...either join me in analyzing the data - if you can - or stop questioning my integrity, simply for insisting on analyzing it properly."
<Sounds like a bit of the "victim" here (bolded), with a dash of implied (attempted?) intimidation thrown in (underlined).>


"-Argument by Rhetorical Question. "Who can be against consensus?""
<Either you're NOT paying attention(?), or you possibly missed the point of this fallacious argument concept. (You're not trying to mislead, are you? ;) ) I wasn't arguing for the idea of "consensus", I was pointing out how using such an obviously non-controversial issue (like "I'm against pollution! (who isn't?)) can be used to mislead people.>

"-Inflation of conflict. Ron and I disagree, you like Ron, therefore I am bad to disagree with Ron because I risk discrediting USCV by having the audacity to pursue the truth in public."
<WOW! again. What a lot of assumptions. I don't "like" or dislike Ron. There ARE things I do like about him though. There IS quite a difference in the two statements. For example, he is the most qualified commentator on this issue here (PHD in statistics), etc. But, if he did something that was counter-productive to advancing election fraud discovery and election reform, I would have no emotional attachment ("liking him") standing in the way of necessary criticism. You are "bad to disagree with Ron"? My my, now isn't THAT a good example of assumptive AND emotive language? ;)"having the audacity to pursue the truth in public"? (The "victim" once again.)>

"-Needling. For example, your suggestion that those who disagree with you should check a website and brush up on their rhetorical skills, with the clear implication that you already posess those skills in abundance."
<Actually, it was a good idea; but it appears you REALLY missed the point. The list is not only meant to help you spot others fallacious arguments; but also YOUR OWN. In other words, it is meant to increase your objectivity (critical thinking skills).

I won't ask you to apologize for your computer skills anymore than you should ask me to apologize for my abilities. They have proven very useful to see through manipulation (and hidden agendas), and are likely to do so in the future. And, of course, they do help me to "check" my own thinking/arguments.>

"-Changing the Subject. You can't criticize my thesis, so you criticize my approach."
<Man! YOU are really a creature of habit, aren't you? ;) You assume I, "...can't criticize <your> thesis, so <I> criticize <your> approach"?Now, really, doesn't that seem like a bit of a "stretch", even to you?

"and last, but not least, my favorite:"
<WOW! You're really on a roll now, huh? ;) >

"- wait for it-"
<Oh, you seem SOOOO proud of yourself. ;) >

"Misunderstanding the Nature of Statistics.

As I said before - and you ignored - I'm so concerned that the election may have been stolen, I'm actually trying to find the evidence. If you've found the evidence, show me. If you haven't, either join me in analyzing the data - if you can - or stop questioning my integrity, simply because I insist on analyzing it properly."
<Now aren't you glad you used this again? ;) I "lifted" it from your other post for the example above. I really didn't realize you used it again down here since I didn't read all the way to the end; and I took your points on "as they came". See above, if need be.

I will add this though. I really wonder if YOU understand "the Nature of Statistics"? Specifically, the use, and limits of, inferred data.

This will be one of my major criticisms of your first post (the "fatally flawed" one). When I first started researching on your post on Wednesday, your use of the concept "proof" ("proven" applying to both rBr and election fraud) really jumped out at me. In contrast, I noticed Ron used terms like "provide support for" in similar circumstances.

Now, I've also noticed that a significant portion of Ron's response to you attempts to paint what can be expected of inferred statistics in more realistic terms: for example, "...varying patterns of partisan exit poll response require "further explanation"...".

Frankly, I think Ron clearly understands the limitations of inferred data - I'm skeptical that you do.>

EDIT: It looks like I missed one. The statement you seem sooo proud of (that you've used it twice), is a "victim" of the notorious "excluded middle".
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#374751

[mgr]

64. Good for the goose, is good for the gander

Nice argument from faith. Having not read Bruce's paper--

"<WOW! again. What a lot of assumptions. I don't "like" or dislike Ron. There ARE things I do like about him though. There IS quite a difference in the two statements. For example, he is the most qualified commentator on this issue here (PHD in statistics), etc. But, if he did something that was counter-productive to advancing election fraud discovery and election reform, I would have no emotional attachment ("liking him") standing in the way of necessary criticism. You are "bad to disagree with Ron"? My my, now isn't THAT a good example of assumptive AND emotive language? ;)"having the audacity to pursue the truth in public"? (The "victim" once again.)>"

When does a Ph D in statistics make one more qualified than a Ph D in engineering, or biology, or psychology? It is the power and substance of the argument which matters, all a Ph D implies is that one has presumed mastery of a small area of arcane knowledge. It does not translate to overall mastery, or authority.

You do a very good job of attacking the form of arguments, but you don't seem able to address the substance. Your points are little more than distractions. Read the damn paper.

Mike

Mike

[mgr]

59. Straw man of a straw man

And the straw man is what? Exaggerating the approbation that anything associated with Mitofski is held on this site?, or is it an infinite reduction since I exaggerate an exaggeration ala Howard Campbell in Mother Night? Or are you engaging in an argument from authority? One can slice an argument anyway one wishes with this laundry list. It is whether one applies it equally to all arguments and positions is the knock.

There is very little, besides innuendo, to come to the conclusion that NEP is inherently a part of the cons(p)iracy to steal this election. What is true is that NEP is a business operation, and must consider its competition and standing when addressing our concerns. I have already argued that compared to other exit polling concerns, their behavior is more transparent and open. The fact is that we should work within that constraint, rather than bash them for no good reason other than to have a villian.

I can see you have never dealt with Friends. You question my use of the term consensus in context of Bruce O'Dell's words. I consider consensus a process that continues until all parties involved either come to agreement, or acknowledge they are at an impasse. I don't see that either outcome has arisen. It's still in the discussion and debate stage. The decision to go forward with a minority paper does not preclude later consensus, but it does place those involved into difficult relationships, that they must engage each other to resolve. And it appears to be a little public, and its on our stage. Big Deal.

Mike

(edited for spelling)

[Time for change]

30. No he is not advocating Mitofsky's position

When statisticians and mathematical modelers use the words "produce results consistent with..." all they are saying is that the case hasn't been proven against the position they are referring to.

Here is a link to the original USCV report, which Bruce O'Dell signed, which strongly criticizes Mitofsky's report: http://exit-poll.net/election-night/EvaluationJan192005.pdf. I doubt that he would want to retract any of what is said in that report. What he is doing in this thread is simply saying that USCV has gone too far by saying that they have proven the case against rBr.

With regard to my motives, as I said in my original post to this thread, I think that this is an important issue, from which we can all benefit by discussing it. You might note that in my first reply to this thread I request a response from Bruce to my reasons for believing that rBr is implausible: http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=374482&mesg_id=374502&page=. In that response I refer him to posts where I pose arguments against rBr, and I ask for his opinion. I have also pm'd him about the same thing. I think that we all need to understand this better, and I think that discussions such as this can lead us in that direction. That is what I am trying to do.

[TruthIsAll]

32. "Fatally flawed" was a comment that GWB used a lot, if you recall

The economic study is "fatally flawed".
The Global Warming study is "fatally flawed".
The intelligence was "fatally flawed".
The lock-box concept is "fatally flawed".
The UN inspectors work is "fatally flawed".


Brings back memories...of the "fatally flawed" resident in the WH.

[Bruce ODell]

44. Did you have anything to say vaguely related to my paper?

...Just curious.

[Name removed]

65. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Melissa G]

85. Thanks for your all your work!

Am having a killer work weekend right now so it will be a couple of days before I have the pleasure of properly enjoying your efforts but Welcome to DU. :toast: :hi:

[TruthIsAll]

141. Your choice of words was incendiary.

You have a lot of chutzpa to call the work of those who have worked with you "fatally flawed" in a banner headline no less.

No class.

Your judgment in using that term is "fatally flawed".

I have read your paper.
And you don't put a dent in Ron's analysis.
He is MUCH more qualified in statistical analysis than you or I.

But, you are a modeler, a systems architect, right?

Then why don't you give me some inputs for my constrained optimization model? This way we can proceed to debunk the fiction that is rBr, especially now that you are a newly minted member of the confounding rBr crew.

I'm sure you would love to do it.
Well, I have created this nice little optimizer to do just that.
It could help.
It can't hurt to stress test it.

Oh, I forgot, you don't believe we can EVER disprove rBr.

But what the hell, let's try anyway.

Now, as for those inputs...

[Bruce ODell]

34. With all due respect, please read my paper

If you actually had, you would have encountered these words:

"...a closer look at the data cite in their report reveals that Kerry and Bush supporter exit poll response rates actually did not vary significantly by precinct partisanship . Systematic exit poll bias cannot be ruled out as an explanation of the 2004 Presidential exit poll discrepancy – nor can widespread vote count corruption . The case for fraud is still unproven, and I believe will never be able to be proven through exit poll analysis alone.

" This paper should not be misinterpreted as an argument against the likelihood of vote fraud. Quite the opposite; I believe US voting equipment and vote counting processes are severely vulnerable to systematic insider manipulation and that is a clear and present danger to our democracy . I strongly endorse the Working Paper’s call to implement Voter-Verifiable Paper Ballots and a secure audit protocol, and to compile and analyze a database of election results."

Perhaps it's old fashioned of me, but working as I do in an engineering profession, truth actually is more than just a peripheral concern.

When the question at hand is: "am I still a citizen of the American Republic, or am I just a serf of the American Empire", I for one strongly believe that getting the answer wrong - either way - is a Really Bad Thing. Don't you?

Has it occurred to you (as I also stated in my paper) that "It is possible to speculate that if a hypothetical group or groups actually have the power to covertly shift votes on a national scale sufficient to change the election outcome, they doubtless would be extremely careful to try to avoid detection – especially, one would think, by angry statisticians."

I'm so concerned that the election may have been stolen, I'm actually trying to find the evidence. If you've found the evidence, show me. If you haven't, either join me in analyzing the data - if you can - or stop questioning my integrity, simply for insisting on analyzing it properly.

[Time for change]

41. I can't link to your paper

Are you sure it's the right link, or that it's still there?

[Bruce ODell]

43. Location of my paper

www.digitalagility.com/data/ODell_Response_to_USCV_Working_Paper.pdf

still works fine for me.

[OnTheOtherHand]

45. just to add to Bruce's link...

Here is a tinyURL variant that should survive copying and pasting, unlike the original: http://tinyurl.com/8mtfe (which redirects to the URL Bruce just posted).

It would be great if DU could relax the length restriction on URLs -- broken links mess us up an awful lot!

[Helga Scow Stern]

36. That was my sense too.

But I will read the paper and see what I think.

[Peace Patriot]

66. If the Bush Cartel had wanted an honest, verifiable, election...

...why didn't we have one?

[Time for change]

67. Obviously they didn't want one

That's why:

We had voting machines with secret software programs for counting our votes.

Every effort is being made by Republicans in Congress to prevent meaningful election reform.

Every effort was made in Ohio (successfully thus far) to prevent a hand recount of the vote.

And

The MSM almost totally ignores this issue.

[Verve]

72. I think it's promising that there is open debate about this. However,

after hearing that at least 1 statistician has received monetary benefits from Mitosofsky (and whomever else for that matter)I really don't know who to believe.

Is this just another way to spin and dis inform the public or is there true academic debate going on here? For non-statisticians who cannot fully digest this material, who are we supposed to believe?

I mean no disrespect to any statistician that posted here with a good conscience. If there really are flaws in these papers than they need to be discussed and debated. That's science!

Yet, if some statisticians are getting paid by the company that is at the core of this debate, that's fraud!

[Febble]

74. Look let's get this clear.

I did all the work I posted here completely independently of Mitofsky. He had absolutely nothing to do with it, and I received absolutely no monetary benefit.

In fact I could have earned some money doing a bit more useful teaching if I hadn't been obsessing with these exit polls.

What my work did was to show that the metric by which E-M had analysed their data was flawed, and that their analyses as to the causes of the exit poll discrepancy was therefore flawed. And, unfortunately, and analyses based on stuff presented in the E-M report, including the USCV report was also, if not flawed, based on flawed input.

In other words, it was back to the drawing board for everyone, including E-m

As a result of my work posted on the internet, for all to see, in which I drew attention to a flaw in the way E-M had analysed their data, Warren Mitofsky contacted me, firstly to say they had taken note of my paper and that he'd done some new analyses, using my formula instead of theirs.

Which was great.

Then, shortly before Mitofsky was due to present this new analysis at the big AAPOR meeting in Miami, he engaged me to do some further work on the data.

That is IT.

The implication in your post is that somehow I was bribed by Mitofsky to say stuff that suited him. Nothing could be further from the truth. I criticised his work. He agreed with my critique. He reanalysed the data using my measure. He presented the findings. He engaged me to do some more.

I have really had enough with these totally unfounded attacks and innuendos.

Flaws need to be aired. They were. Mitofsky noticed. He was interested enough in the flaws in his own work to bother to reanalyse it. For all I knew it might have proved fraud.

For all I know it might yet.

[Name removed]

78. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Time for change]

76. I wouldn't have posted this thread if I wasn't pretty sure

it was true academic debate.

If you look up other of my recent threads you'll see that I feel pretty sure that John Kerry won the 2004 election. And because of that I was somewhat reluctant to post this, but I think that this debate is a good thing, because we need to continue to search for the truth in order to better understand what happened, so that we can expose what we find.

It is important to understand that the two main participants in this debate have played prominant roles in U.S. Count Votes (USCV), an organization devoted to analyzing the exit polls in order to try to ascertain what went wrong in the 2004 Presidential election, and to advocating election reform so as to prevent election fraud in future elections. Somebody tell me if I'm wrong, but I'm pretty sure that neither of them has received a penny for the great amount of work that they have done for USCV -- and even if they have, I certainly wouldn't hold this against them, because this is a very important cause.

Furthermore, what a lot of people don't understand is that this debate is not about whether John Kerry really won the election or not. One side thinks that the USCV analysis of the exit polls prove, or almost prove that he won, whereas the other side thinks that the exit polls don't prove the case either way. Both sides are strong advocates of election reform and continuing to analyze data relating to the election so that we can understand it better.

[Name removed]

73. Deleted message

Message removed by moderator. Click here to review the message board rules.

[autorank]

80. Bill Bored, This One's for You

:rofl::rofl::toast::rofl::rofl::toast::rofl::rofl::toast::rofl::rofl:

[Bill Bored]

105. I posted a JOKE, and it's gone!

I guess these exit poll threads are not to be made light of in any way, shape or form, even if it's made perfectly clear by the poster that s/he is just trying to inject a little humor into the process.

Strange when one considers that both sides of this debate now seem to admit and agree that the exit polls can not, on their own, be used to prove fraud. (See post# 16.):

O'Dell said:

"The case for fraud is still unproven, and I believe will never be able to be proven through exit poll analysis alone."

Baiman said:

"I agree."

Thanks to both of you for finally putting this issue into perspective.

Apparently it's OK to have an endless debate about a methodology that can't even prove election fraud, but it's taboo to make a little joke about it?

Well, I only hope that those of you working on this problem can devote even half as much time to developing other evidence.

If not, I must reiterate my stance that this is a huge brain drain, and unfortunately, THAT"S NOT A JOKE.

[autorank]

106. Bill Bored, you know I loved that humor!

NEP's are a wonderful pointer to the stinking cauldron of election fraud. So's the Conyers Report Great Thread on Conyers What I've noticed is that this thread is primarily the recently arrived academics plus people questioning the process. Thus, for the academically inclined, these threads seem a bit self-stimulatory (if ya know what I mean).

I like TIA's work because it's just obvious, and a key indicator that fraud took place. I like the other aspects of election fraud work, which necessarily join the high-level stuff. Any good prosecutor would look at everything that's out there, citizen gathered and developed, and say, let's get ready to rumble.

With regarde to your previous question about relationships, I'm sure that these folks will have the answer somewhere http://www.familysearch.org/ (A wonderful resource and public service)

[Bill Bored]

114. Here's the deal

I have been saying for some time that the oddest thing about the national exit poll was the need to readjust party ID weightings from 38 Dem/35 Repub to 37/37 to get the numbers to jibe. This is what put Bush in the running; otherwise he would have lost. Pure and simple.

It should be a simple matter to determine if this change in relative party turnout actually occurred, or if actual party registration reflects it. The effect was strongest in Mitofsky's "Western Region", whichever states those actually are, in which the Party ID percentages were actually reversed on 11/3 compared to the night before! How could this be? How could the Dems have stayed home in 2004? Or how could there be so many Republicans, and all voting for Bush??? Please! I've never once even set foot on a turnip truck, much less fallen off one recently!

I respect the work of everyone involved, but we have to step back and look at the bigger picture. TIA has been the only one to seriously look at the Party ID anomaly, even though, unlike many other aspects of this stuff, this is something that can probably be checked independently of both the exit polls and the official count.

Anyway, I think it's much more interesting to look at e-Ballot Definition programming. That's the way to steal an election in this country today, and there are lots of examples reported in the press about how screwed up it is, even if the authors of those stories don't grasp their full significance. I expect more to come out about this as time goes by.

The resignation of the Miami-Dade Elections Supervisor is but one recent example of this. So why isn't ES&S being held accountable? They set up the machines, she didn't. Could it be that we are all too focused on exit polls to really care? The taxpayers of FL may be out $25,000,000 too. Oh, and you know, FL was even outside Mitofsky's MOE. ;)

[LightningFlash]

118. Yes this is an issue.

Now that the United States cencus data is officially released, I believe it is time to thuroughly analyze that very issue. The more time we spend wasted on strawmen arguments about "the proper" measuring tool for fraud, the more time we waste catching the actual criminals in Florida.

Those ES&S employees paid by the BOE are laughing at us, not because they simply got away with it....But because we are too wrapped up in meaningless, pointless, and stupid "fancy functions" than to actually look their direction.

;)

[Febble]

156. I completely agree n/t

[tommcintyre]

115. "these threads seem a bit self-stimulatory (if ya know what I mean)."

You mean sorta like a "numerical circle-jerk"?

Possibly satisfying at the moment, but NOT very productive.

:rofl:

[tommcintyre]

82. OK... is it timefor me to put on my gum shoes and Fedora? ;)

B-)

[Name removed]

88. Deleted message

Message removed by moderator. Click here to review the message board rules.

[autorank]

86. Time for change, You Started Strong but You're Finishing Weak!

I did like you "non statistician" piece. Nice breath of fresh air.

I am profoundly disappointed in this thread and your participation in what I see as either odd contributions or publicizing internal squabbles on DU Elections.

An Odd Contribution:

Febble and O'Dell are numbers folks (statistics, research, whatever). They both say, in slightly different ways, that they are really concerned about fraud in election 2004. Then, in what is a peculiar twist, they both proceed, in desultory ways, to seriously question the statistical evidence for fraud. Does this seem odd to you? If my main concern is election fraud, I'll go out and work on that. However, if I say my main concern is election fraud and I use my specialty to seriously question the use of my specialty in election fraud work (febble's rBr; O'Dell's petulant memo featured by you), then the question is WTF is going on. rBr is a marginal piece of the puzzle, demolished completely here, and O'Dells intramural conflict in just one group of many fighting election fraud is grandstanding.

Publicizing Internal Squabbles:

And then there is the O'Dell memo posted by you. Just above you said "I wouldn't have posted this thread if I wasn't pretty sure
it was true academic debate." Well, please tell me how a memo from a private corporation, released on a political forum fits in the category of academic debate. IMO, the memo is incendiary, it divides one of the key groups in the tiny election fraud movements, and it's featured here, which adds further controversy. Once again WTF. You mentioned that you were a professor. At your university, does the faculty engage in academic debases by releasing incendiary press releases on the campus' local area network. I don't think so. Until you made this statement, I hesitated to see this post as a divisive move. Now, it would appear, I should re examine that position.

Finally, for those interested in election fraud and the positions of the commentators around here, this is a response by OnTheOtherHand to a question I posed: "What's your most compelling case for the bias and how did it influence the 2004 elections?"
---------------------------------------------------------
An OnTheOtherHand post on election fraud

"I'm not sure what I think, about the 2004 election. I think Bush probably won the popular vote (which leaves room for voter suppression, as we know happened). But I'm not arguing it, because there's a lot I don't know. I can explain the reasons why I'm not convinced that the exit polls prove massive fraud, but that doesn't mean that I am ruling out massive fraud. And since massive fraud is possible, we have a big problem whether or not it actually happened in 2004, so in a way I don't even care that much whether it did.

I bet someone is gonna misread that and jump down my throat. But you asked a fair question, and that's my best brief shot at answering it right now"
-----------------------------------------------------------

"Academic politics" makes for strange bed fellows!

NEW LEADERS FOR A NEW DEMOCRATIC PARTY!

Contact the DNC and Give 'em Hell About NOT Acting on Election Fraud


This is what it's all about, political action, not paralysis by analysis.

[Febble]

89. You say:

Febble and O'Dell are numbers folks (statistics, research, whatever). They both say, in slightly different ways, that they are really concerned about fraud in election 2004. Then, in what is a peculiar twist, they both proceed, in desultory ways, to seriously question the statistical evidence for fraud. Does this seem odd to you?

This seems normal to me (apart from the "desultory" part, which is simply wrong). Numbers folks on the whole listen to the numbers. Good statisticians, despite the adage, do not lie with statistics, or they try not to. They try to tell it like it is. A paper appeared that claimed that something was "not consistent with the evidence". That is a strong statement. Bruce and I, as numbers guys, concluded mostly independently that the statement was not supported by the numbers.

Well, please tell me how a memo from a private corporation, released on a political forum fits in the category of academic debate.

.

This is how it fits into an academic debate: USCV have produced a number of papers as a body. These have been widely, and rightly, regarded as important. Three things have given them a credibility normally reserved for peer-reviewed academic papers:

Many of the authors have PhDs (a fact pointed out in press releases)
They were published under the banner of an organisation committed to Election Reform
The organisation boasts a "core group" of "statisticians" who "review" the published work.

The pieces were taken seriously and applauded by many on DU. They fitted the DU agenda: political; related to election reform. So far so good.

Then it emerges that the review process and the consensus claimed for USCV, are not as solid as they seemed.

Suddenly USCV becomes "a private corporation", and Bruce's whistle blowing is "a memo".

Academic debate does happens in the open, mostly. Peer-review happens in private, but it is, crucially independent, as are the editorial decisions based on the reviewers' recommendations. The USCV working paper had no independent peer review and no independent editor. It was released, citing O'Dell's work, but without his imprimatur. One author claimed on DU that reason it did not have more "authors" (a term used by USCV for assenting internal reviewers as well as writers) was that other authors had been "busy". Two authors also claimed, on DU, contrary to the timeline evidence, that I had plagiarized USCV's work.

O'Dell has put the record straight. If you want to use uncritiqued, and possibly flawed work for your cause, you are welcome. But to maintain that exposing its flaws, both in substance and process, is not relevant to the debate seems, to me, to use your word, "odd".

[autorank]

99. I knew I was on to some thing with"desultory". Academic debate?

How does the head of a research firm, private company/corporation, publishing a a memo about a non profit on a political forum constitute "academic" debate. We're a little far from the Academy here aren't we.

[Febble]

102. OK then

why does a memo from a member of that non-profit (Ron Baiman) constitute a contribution to the debate? Whereas Bruce's doesn't? When Ron's conclusions were based on Bruce's model?

If you don't want to debate either of the USCV reports because you think they are a distraction, say so.

I agree with you that the entire exit poll debate is a distraction. It's a point I've been making for some time. In a sense it was the point of my original paper. It has certainly been the point of most of my DKos posts.

But I didn't think that was your view.

[autorank]

104. So why don't you go here (with your friends) and debate the point

http://upload.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374958

...btw, this is a rhetorical question. No need to respond. Your response will be judged on whether or not you start engaging TIA on his recent work rather than lounge in these "academic debates."

How about a real debate?

[TruthIsAll]

108. Febble, The exit poll debate is NOT a distraction; rBr is.

And that is the BOTTOM LINE.

A distraction? Because you say so? Didn't you also say that "mounting evidence" will show that Bush won the election?

Remember that? When you made that statement you lost me forever. It was a clear giveaway. And that was before we found out that you were working with Mitofsky.

THE EXIT POLLS HAVE SHOWN ALL REASONABLE AND KNOWLEDGEABLE PEOPLE THAT KERRY INDEED WON BEYOND A REASONABLE DOUBT. THAT IS THE REASON FOR ALL THAT ANALYSIS.

NOW LETS GO FROM THERE AND MOVE OFF THAT FAITH-BASED, PABLUM BULLSHIT WHICH IS CALLED RBR. IT HAS BEEN PROMOTED AND SOLD TO THE MATHEMATICALLY CHALLENGED PUBLIC FOR ONE REASON ONLY- TO CONFUSE AND HIDE THE TRUTH. BUT TRUE ACADEMICS KNOW IT'S A FARCE.

AND THE AMERICAN PEOPLE MAY LACK MATHEMATICAL MATURITY, BUT THEY KNOW WHAT A DEAD RAT SMELLS LIKE. SO ITS ONLY A MATTER OF TIME. WE WILL GET THERE WITH OR WITHOUT YOU.

I MUST ADMIT, YOU GUYS ARE TENACIOUS.
SO ARE WE.

[Name removed]

109. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Febble]

116. Please point to the post in which I said that

And I was not "working for Mitofsky" when I did all work that has been discussed on this site. It was because of this work that Mitofsky engaged me to run some more analyses.

Please read my other posts on this subject.

I would be happy to debate with you. I can't debate with someone who roars abuse and makes unfounded attacks on my integrity.

I think the exit polls are a distraction because my reading of the numbers does not convince me, as it clearly convinces you, that there was massive fraud. I think there is much stronger evidence available. And I think there is an irrefutable case for demanding fair, secure and auditable elections.

But I am happy to debate the numbers. I am not happy to be abused and insulted.

[TruthIsAll]

117. Abused? Insulted? Yes, you said it about a month ago.

I do not have the time to go back and look for it.
But you said it.
About one month ago, if not here than on dKos.

I mentioned it once before.
But you conveniently forgot.

I'm sorry if you take this as an insult.
It is not abuse.
The truth is never abuse.

BTW, do you have some inputs for the EXit Poll Response Optimization model?


http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x375074

[TruthIsAll]

119. OK, FEBBLE, YOU ASKED FOR IT SO...THE TRUTH

But I think the evidence is mounting that Kerry did not win the popular vote, and that the Great Exit Poll Discrepancy was largely due to poor random sampling. I realise most on this forum disagree. I can only say that I have not come to this view through bias. I was massively biased in favour of finding the opposite.

Thanks for having me. I share your pain.

Lizzie

(Elizabeth Liddle)

YOU SHARE OPUR PAIN?


http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=366252

[Bill Bored]

122. Febble see post 114

Here's the link:
<http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=374482&mesg_id=375093&page=>

Can Warren explain why they had to readjust party IDs so radically, esp. on the Western region? Is there no way to see what the true turnout was among Dems and Repubs? How could these particular weightings have been so far off, especially in the West?

It was well known that there were Republicans for Kerry in 2004, even conservatives (because Bush is truly no conservative). Yet in order to get the polls to match the count, they had to depress the Democratic weighting and embellish the Republican one.

So where did all these Republicans come from, and what happened to the ones who voted for Kerry, and why did the Democrats apparently stay home, particularly in the West?

Could it be because the exit polls were released early and showed Kerry had it all locked up? Or is the whole thing just Warren's best guess to explain away the possibility of vote count corruption?

Frankly, I don't give a damn about reluctant responders! I care about the fact that these polls say my party didn't even bother to show up to vote and this after the biggest grass roots campaign effort in history! That's the issue. (And also that, unlike the Warrens of the world, I KNOW just how easy it is to swing elections on GEMS servers and the like.)

The burden of proof should be to show that this did NOT happen. If this exit poll data can help, then fine, otherwise it's junk.

I would like an explanation for the party ID thing though. That is quite unprecedented and unlike some of the other stuff being looked at, it sticks out like a sore thumb.

[LightningFlash]

123. Also, the official Bureau of Cencus statistics shows the full page.

http://www.census.gov/Press-Release/www/releases/archives/voting/004986.html

In fact, if we are to go by traditional means than the measurement offered by the U.S. cencus bureau is far more accurate than a manufactured measuring tool, by one who is a non-PHD and those who are accredited PHD's as well.

[Bill Bored]

139. Nothing about Party ID in there.

But some of the other stuff could be compared to the exit poll to see how far off either the pre-adjusted or adjusted versions were.

TIA may have done something with this already, but it would have to be a pretty rainy day for me to get into it. If the female or youth vote turned out to be stronger than in the exit polls, it could yield something, but it's the party ID stuff that really had to be messed with for Bush to win.

And I suppose this is consistent with rBr stuff. Warren says the Dems didn't show up to vote, but the ones that did took the exit poll. OK, sure.

[OnTheOtherHand]

168. the problem with party ID is

that it isn't the same as registration.

In the national exits, people were asked, "No matter how you voted today, do you usually think of yourself as a: Democrat, Republican, Independent, Something else." We know from panel surveys that people's responses to that sort of question aren't very stable, and they can't be reliably checked against registration data -- the balance of "Independent" Democrats to "Independent" Republicans is unknown and probably also unstable.

I'm not saying that you should accept the exit poll weightings -- on the contrary, they are basically made up after the fact to match the official returns. Not a cover-up (IMHO), just standard practice (not my opinion). And I like the idea of seeing how many registered Reps and Dems voted in, say, Ohio -- it might not be conclusive, but it could be pretty telling.

[Bill Bored]

193. Yes but I don't think that question was asked in the state polls.

I think the issue may boil down to this:

Warren says there were reluctant Bush responders. But his final weighting says there were reluctant Kerry voters! In other words, he adjusted his poll to indicate that the Kerry voters stayed home, relative to prior years' Democratic voters. This is an anomaly that ought to be investigated.

For the results to match the rBr hypothesis, Warren has implied through his final weightings, that Kerry voters were reluctant to come out and vote for him. This is not a reluctant responder hypothesis, it's a reluctant VOTER hypothesis.

If he had produced a final version of the poll that showed party IDs comparable to those of the past, the rBr thing might hold more water. But he did not. Instead he just implied that Kerry voters stayed home, or that Bush voters showed up in droves.

Rather than beating this reluctant responder thing to death, why not try to find out if there were really reluctant Kerry voters? This is what's implied by Warren's final weighting, not the ones used on 11/2. It is his version of the "truth" as expressed by the final weightings and the vote count.

By suppressing the number of respondents who identified themselves as Democratic, he's in effect saying, Bush won because Kerry voters stayed home -- NOT that Bush voters didn't respond to his poll. And by focusing so relentlessly on rBr, USCV and the rest are letting him get away with it.

Remember, this is the final weighting, which is supposed to take all that response bias stuff into account, and still Warren is implying that Kerry voters didn't show up at the polls. Sorry, I don't buy it.

[LightningFlash]

204. Check the cencus data by groupings.

Then corroborate it with the polls, you can determine party IDs.

http://www.census.gov/population/www/socdemo/voting/cps2004.html

[OnTheOtherHand]

212. when you say Kerry voters

do you mean likely Kerry voters, namely past Democratic voters plus new voters? It's a fair line of analysis, and an important one regardless of the extent of fraud (and vote suppression, which I think would show up in your line of analysis but probably not in exit poll analysis). It's important because if likely Kerry voters didn't vote for Kerry, and fraud and vote suppression weren't the reasons, Democrats should want to know what the reasons were.

(I'm pretty sure there was some party ID question in the state polls, but I haven't poked around to see whether it was worded differently. We can spot-check a few -- Ohio would be a good place to start.)

[Bill Bored]

217. If you mean Y2K voters interviewed in 2004, not exactly.

Likely Kerry voters, yes, but in the sense that they identified themselves with the Democrats in 2004.

In previous exit polls, the turnout was more Dems than Repubs, which I believe was reflective of overall party registration at the time. TIA has posted some of those numbers. Even when a Republican won the Presidency, there were more Democratic than Republican voters at the polls.

This time, in the final weighting, it was necessary for E-M to adjust the party ID (i.e., the party with which the respondents identified) to be equal nationally (37/37), and more Repub than Dem in the West (38R/35D if memory serves me). Without this particular weighting change, it was algebraically impossible for Bush to win the exit poll. That one weighting change brought him within the MOE. I agree that could reflect voter suppression, but in the West where the adjustment was greatest? Not so sure.

Since almost 100% of the sample answered this question (unlike the "Who did you vote for in 2000?" question), and the answer before the final adjustment was consistent with historical norms i.e., more Dems than Repubs, we have an anomaly in the exit polls. It was never (AFAIK) necessary to change this weighting for the poll result to match the count and certainly not the electoral outcome, and it implies that likely Kerry voters stayed home, or that Repubs came out in droves, and particularly in the West. I think this is unlikely.

Add to this the "Republicans for Kerry", who made themselves quite evident in the media, the Op Ed pages, the web, and so on, and it's even more unlikely.

[Internut]

219. Take a look at this, Bill

http://people-press.org/reports/display.php3?PageID=750

Note that this paper was published in 2003.

[Bill Bored]

223. I answered in the other thread but

surely Mitofsky knew of this work. So why did he release his polls with the higher Dem weighting? On the other hand, I guess it's possible that a lot of new Dems registered in 2004.

The more I look at this stuff though, the more I think these polls are just junk. Sorry guys. The brain drain continues.

[LightningFlash]

224. Becomes irrelevant.

Simply verify the final grouped weightings knowing 1.5 million bush voters since 2000 have died (that's being conservative) and the reversed gender reduction, you have a bogus weighting plus final mix.

"http://www.exitpollz.org/CNN_national1253.html

" 1. VOTE BY GENDER

BUSH

KERRY

NADER

TOTAL
2004
2000
2004
2004



Male (46%)
55%
+2
44%
0%



Female (54%)
48%
+5
51%
0%



2. VOTE BY RACE AND GENDER

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



White Men (36%)
62%
n/a
37%
0%



White Women (41%)
55%
n/a
44%
0%



Non-White Men (10%)
31%
n/a
67%
1%



Non-White Women (12%)
24%
n/a
75%
0%






3. VOTE BY RACE

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



White (77%)
58%
+4
41%
0%



African-American (11%)
11%
+2
88%
0%



Latino (8%)
44%
+9
54%
1%



Asian (2%)
42%
+1
58%
*



Other (1%)
41%
+2
54%
2%


13. VOTE BY EDUCATION

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



No High School (4%)
50%
+11
50%
0%



H.S. Graduate (22%)
52%
+3
47%
0%



Some College (31%)
54%
+3
46%
0%



College Graduate (26%)
52%
+1
46%
1%



Postgrad Study (17%)
45%
+1
54%
1%


15. VOTE BY PARTY ID

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



Democrat (37%)
11%
+0
89%
0%



Republican (37%)
93%
+2
6%
0%



Independent (26%)
48%
+1
50%
1%


17. HAVE YOU EVER VOTED BEFORE?

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



No (11%)
46%
+3
53%
0%



Yes (89%)
51%
+3
48%
0%






18. VOTE BY RELIGION

BUSH




KERRY

NADER

TOTAL
2004
2000
2004
2004



Protestant (54%)
59%
+3
40%
0%



Catholic (27%)
52%
+5
47%
0%



Jewish (3%)
24%
+5
75%
*



Other (7%)
24%
-4
74%
2%



None (10%)
31%
+1
68%
1% " "

TOTAL POPULATION
.Total 215,630 142,197 65.9 73,433 34.1 125,880 58.4 89,750 41.6 142,197 54,869 18,563
..White alone 176,571 120,019 68.0 56,551 32.0 106,687 60.4 69,883 39.6 120,019 42,983 13,569
…White non-Hispanic alone 151,457 111,409 73.6 40,048 26.4 99,662 65.8 51,795 34.2 111,409 36,815 3,233
..Black alone 24,950 16,081 64.5 8,869 35.5 14,062 56.4 10,888 43.6 16,081 7,313 1,556
..Asian alone 9,227 3,227 35.0 6,001 65.0 2,751 29.8 6,476 70.2 3,227 3,001 3,000
..Hispanic (of any race) 27,028 9,309 34.4 17,719 65.6 7,594 28.1 19,434 71.9 9,309 6,756 10,964
..White alone or in combination 179,009 121,624 67.9 57,385 32.1 108,038 60.4 70,971 39.6 121,624 43,671 13,715
…White non-Hispanic alone or in combination 153,451 112,799 73.5 40,652 26.5 100,828 65.7 52,623 34.3 112,799 37,400 3,252
..Black alone or in combination 25,556 16,457 64.4 9,099 35.6 14,374 56.2 11,182 43.8 16,457 7,505 1,594
..Asian alone or in combination 9,658 3,488 36.1 6,170 63.9 2,963 30.7 6,695 69.3 3,488 3,156 3,014
VETERAN
.Total 23,747 18,952 79.8 4,795 20.2 17,367 73.1 6,380 26.9 18,952 4,678 117
..White alone 20,538 16,557 80.6 3,982 19.4 15,208 74.0 5,330 26.0 16,557 3,908 74
…White non-Hispanic alone 19,523 15,840 81.1 3,682 18.9 14,579 74.7 4,944 25.3 15,840 3,643 39
..Black alone 2,369 1,802 76.0 568 24.0 1,628 68.7 741 31.3 1,802 541 26
..Asian alone 273 171 62.8 102 37.2 160 58.8 112 41.2 171 91 11
..Hispanic (of any race) 1,137 796 70.0 341 30.0 701 61.7 436 38.3 796 296 45
..White alone or in combination 20,860 16,805 80.6 4,055 19.4 15,419 73.9 5,440 26.1 16,805 3,980 76
…White non-Hispanic alone or in combination 19,804 16,060 81.1 3,744 18.9 14,766 74.6 5,038 25.4 16,060 3,703 41
..Black alone or in combination 2,447 1,852 75.7 595 24.3 1,669 68.2 778 31.8 1,852 569 26
..Asian alone or in combination 300 188 62.8 112 37.2 176 58.5 124 41.5 188 100 12
NONVETERAN
.Total 191,883 123,246 64.2 68,637 35.8 108,512 56.6 83,370 43.4 123,246 50,191 18,446
..White alone 156,032 103,462 66.3 52,570 33.7 91,479 58.6 64,553 41.4 103,462 39,075 13,495
…White non-Hispanic alone 131,935 95,569 72.4 36,366 27.6 85,083 64.5 46,851 35.5 95,569 33,172 3,194
..Black alone 22,581 14,279 63.2 8,302 36.8 12,434 55.1 10,147 44.9 14,279 6,771 1,530
..Asian alone 8,955 3,056 34.1 5,899 65.9 2,590 28.9 6,364 71.1 3,056 2,910 2,989
..Hispanic (of any race) 25,891 8,513 32.9 17,378 67.1 6,893 26.6 18,998 73.4 8,513 6,460 10,919
..White alone or in combination 158,149 104,819 66.3 53,330 33.7 92,618 58.6 65,531 41.4 104,819 39,691 13,639
…White non-Hispanic alone or in combination 133,647 96,740 72.4 36,908 27.6 86,062 64.4 47,585 35.6 96,740 33,696 3,211
..Black alone or in combination 23,109 14,604 63.2 8,504 36.8 12,704 55.0 10,404 45.0 14,604 6,936 1,568
..Asian alone or in combination 9,358 3,300 35.3 6,059 64.7 2,787 29.8 6,571 70.2 3,300 3,057 3,002

http://www.census.gov/population/socdemo/voting/cps2004/tab15d.xls

Mystery Pollster has failed to out-stand the government. Since the government takes the Census, they know more than him and basically always will.

[OnTheOtherHand]

228. release?

I think it's pretty likely that the final exit poll weightings are dead wrong about something, although I don't have much of an opinion about what (or whether it matters). Weighting after the fact is a speculative exercise. I'm not yet convinced that the party ID is misweighted, but I don't rule it out. However, reweighting the R/D balance (correctly or wrongly) may make more sense than shaving a few points off Kerry's margin among Democrats or Independents, if that is the alternative to match the official returns. Feel free to direct me to other places where you've developed the arguments -- I don't mean to make you retype yourself.

My question here: "why did he release his polls with the higher Dem weighting?" Are you talking about the initial 2004 exit poll release before the forcing to the official results, or something else?

"The polls are just junk" is probably not a bad first approximation -- we are sort of picking through the pieces. But I might as well catch up on what you've found there.

[Bill Bored]

233. Yes there were 2 early releases of the National poll on 11/2.

3:59 PM and 7:33 PM

If you plug the weightings into a spreadsheet, you can see
which ones had to change to change the vote totals in the 11/3
version. The only one that put Bush in the running was this:

"No matter how you voted today, do you usually think of
yourself as a:"

11/2 7:33 PM:
      %          K%      B%
Dem 0.38	0.9	0.09
Rep 0.35	0.07	0.92
Ind 0.26	0.52	0.44

11/3 1:30 PM:
      %          K%      B%
Dem 0.37	0.9	0.09
Rep 0.37	0.07	0.92
Ind 0.26	0.52	0.44

In the first one, Kerry wins by 3.1%.
In the second one, Kerry wins by only 0.6%.
That put the outcome of the election (popular vote) within the
MOE.
All other weighting changes had a minimal effect, but taken
together, they explained the adjustment. But this was the
mother lode.

So I thought it might be useful to see what the party turnouts
were in reality, or at least the party ID turnouts.

No big deal. I've basically had it with these polls, but I've
been saying that since New Year's Eve!

There were also 4 regional breakdowns of the same poll: East,
Midwest, South and West.

FYI, the above weightings changed most in the West from
Dem 37/Rep 34 to
Dem 34/Rep 36

One would think that someone with Mitofsky's experience, given
the Pew report, etc., could have done better.

[autorank]

234. Thanks for the plain English. That's extremel clear.

So when do a "This thread is for autorank" day. It could just be all deleted messages (haha).

[OnTheOtherHand]

235. thanks, good figures

I'm mildly surprised that the West stands out, because from a recent colloquy with TIA it seemed that the exits were furthest off in the East -- but a lot of the huge disparities were in very small states, so they may not show up when the results are summed. (Still, Pennsylvania....) Hmmmm, maybe something to do with my GIS software in August (I'm on the road in July, and June is pretty well shot). Not that I expect any earth-shattering insights, I would just like some more ways of visualizing this stuff.

I sure don't have a clear sense of the information that Mitofsky was getting early evening 11/2, but I think it would have been hard for him to make an experience-based judgment of whether the Dem/Rep split should be closer to +3 or even. If anything, experience would favor the +3, the Pew report would favor the even, and.... At that point he would have no official returns (right or wrong) to check against. Considering that smart people don't agree even now what the right figures were, I don't see how he could have won!

[TruthIsAll]

237. Huge disparities in...

NY, MA, CT, PA, OH, FL, NC...

not exactly small states

[OnTheOtherHand]

238. right, although I'm not sure

which of those CNN counted as "East." NY, surely. OH, FL, NC, probably not -- they didn't go by time zone. (OH, probably Midwest; FL and NC, probably South. But I don't know.)

[TruthIsAll]

239. THE EAST WAS THE BEAST- BOTH REGION AND TIME ZONE

ALL 22 states in the EST deviated to Bush from the exit polls.

13 of them were in the Eastern region.

Do you want me to show you the probabilities?

[Bill Bored]

240. Well, this goes back to New Year's Eve

when someone got a copy of the regional breakdowns and posted links to them here. It cost him a pretty penny too!

The one thing we don't know is which states are in which regions, however, by using the N of respondents in each region, and comparing them to state populations, EVs, etc., it might be possible to do a fit test to correctly allocate any borderline states. (We know the total N from the national versions produced at the same times.)

There were actually a few attempts at this.

The reason I think it's interesting is that it implies a causal relationship between the exit poll adjustment and the party turnout, which I was hoping could be verified independently of both the exit polls and the vote count, which is after all what we need. In other words, it represents Mitofsky's best guess on 11/3 as to what happened on 11/2. All this rBr stuff came out later, at least publicly.

But no one has really picked up on this and I'm much too busy with other stuff. Exit polls are not my priority.

If anyone wants to take a crack at it, I can send links to the PDFs, if they're still out there. I think they are, in much the same way Diebold software still is. (You just have to know where to look.) So PM me.

[OnTheOtherHand]

241. not yet, but here is a candidate for first crack at the region defs

ftp://ropercenter.uconn.edu/United_States%5CVNS/USVNS2000-NATELEC/version2/usvns2000-natelec.pdf

page 17

I probably "should" have the corresponding 2004 code book somewhere, but don't see it. Not my priority either, although I agree that it's interesting. (Yes, Bill and I have strange interests.)

[Bill Bored]

243. This agrees with TIA's regions.

Might be a way to detect massive voter suppression as you've suggested. We tend to think of the West as CA, OR and WA, but NV is in there and we know there were dirty tricks there such as telling the Dems to show up on the day after the election. Hey I wonder if they conducted any Exit Polls on 11/3? THEY would have had a high Dem weighting, wouldn't they? ;)

And Kerry lost NM and NV by only a few thousand votes.
Per Febble's "other" report, there were a lot of undervotes in NM's minority/DRE precincts. Could have easily been caused by corruption of Ballot Definitions there. And don't forget Iowa, but of course that's not in the West. Only a 10,000-vote margin for Bush in IA though!

[TruthIsAll]

242. BILL: HERE ARE STATES/REGIONS - THEY CONFIRM THE 13047 NATIONAL POLL

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x362208

[mgr]

236. Which time zone? n/t

[kiwi_expat]

218. Party ID question on NEP's Ohio questionnaire

The following NEP Ohio RAW (unweighted) numbers are from 
Blue22's Ohio spreadsheets of the Univ.Mich. NEP data.

Vote for President?

Kerry             1092
Bush               924
Other                4
No vote/No answer   22


Party ID?

Democrat    761
Republican  678
Independent 482
Other       121

--------------------

Blue22 also calculated the percent
 of each Party ID who voted for KERRY:

Democrat    90.7%
Republican   5.8%
Independent 63.1%
Other       48.8%


(Blue22 has kindly offered to e-mail his spreadsheets to
anyone who PMs him their e-mail address.)

[Febble]

166. All I have analysed

is the precinct-level error.

All error sources would have contributed to the extent of the re-weighting, including precinct sampling and presumably demographic weightings, and thus to the discrepancy between the early and late estimates.

But the precinct-level error was very great. This error sources has nothing to do with weightings at all - it is simply the difference between the proportions of each type of voter (Kerry, Bush) sampled, and the proportions of votes cast.

And there is a very large discrepancy at this level (larger than the error in the early state estimates). For some reason, or reasons (almost certainly plural), at an absolutely basic, pre-weighted level, more Kerry voters per Kerry vote were polled than Bush voters per Bush vote. This could have been because Kerry votes were lost/switched, or because Kerry voters were more likely to respond to the poll. We need to know which, and one way to try and find out is to have a) a good measure of the extent of the bias at precinct level and b) regress this measure on multiple factors, including machine type and precinct partisanship. I totally agree with Ron Baiman that this is what is required.

Investigations into the party ID thing you suggest may also help. My main reason for doubting the polls was the massive turnout, which, as you say, ought to have favoured Kerry. When I saw those lines in Ohio, I thought he'd won.

BTW, I got your joke! I thought it was quite funny actually, but only if you knew it was a joke (and knowing your posts, I thought it was). But after the connection I have seen made between me and the RNC on this forum (via the fact that someone in my psychology department has a Wellcome Trust grant, and Wellcome is connected to Glaxo, and Glaxo supports the Republican party....) I worried it could be misinterpreted - and it seems someone did.

[Bill Bored]

215. Besides, in the post, I said it was a JOKE!

Glaxo are also bad news because they threatened to withhold their drugs from Canadian pharmacies that sold them to Americans at reduced (relative to American) prices.

At least they don't make votin' machines!

[TruthIsAll]

135. I pointed it out - and still no explanation from you

Mulling over how you are going to spin that statement away?

Just what brought you to the conclusion that Bush won the popular vote? Does that not indicate a bias on your part against the polling results?

What is your rationale for calling it biased polling? Did you just get careless and spout some future talking points?

Where is your evidence?

The evidence for fraud abounds.
86 of 88 touchscreens switching Kerry votes to Bush.
Ohio
Florida
New Mexico
North Carolina
Nevada
Virginia

etc.

If you knew anything about our voting history, you would know that if Ohio went for Kerry it means that he won the popular vote easily, because NO republican has EVER been elected without carrying Ohio.

Febble, it's very clear.
Your interests lie in serving your current (and future) benefactors, not in restoring our former democracy.

[Febble]

142. Give me a chance!

I asked my question last thing before going to bed, I check in first thing this morning - you know I live in the UK, because it is one of the things you and others have used to disparage my opinion.

Yes, I think the evidence is mounting that Bush did not win the popular vote. You said I said the evidence is mounting that he won the election.

I don't know about the election, because I don't think Ohio has been properly investigated, and there was no recount. It was a sham. You have an electoral college in the US. I also think there may have been widespread instances of vote-corruption of various kinds. I also think the fact that the US election is neither secure nore auditable is a scandal, and does raise serious questions about whether Bush won the electoral college vote. But evidence for the case that Bush won the popular vote comes from the exit poll evidence, and I think that evidence is mounting the magnitude of the exit poll discrepancy is unlikely to be attributable, in its entirety at least, to fraud.

Thanks for the reference to my comment - it makes a difference to know exactly what I said.

I stand by the statement I actually made.

[tommcintyre]

143. Febble, I think a lot of us are having problems with you regarding...

things like the following in this post.
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#375117
I have always had hopes you could "break through" this cognitive dissonance thing. Why don't you read this post, and address it REALLY OBJECTIVELY please.

Also, please feel free to address this one to OTOH. It is also at the "heart" of the problems. (I've been thinking of addressing it to you and Bruce ODell also.)
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#375168

Thanks

[TruthIsAll]

189. FEBBLE, THAT IS JUST THE OPPOSITE OF WHAT YOU SAID!!!

YOU ARE PLAYING WITH WORDS.
YOU ARE SPINNING, GYRATING MADLY.

YOU SAID EVIDENCE WAS "MOUNTING" THAT KERRY DID NOT WIN THE POPULAR VOTE.
YOU DID NOT SAY ANYTHING ABOUT THE ELECTORAL VOTE.

YOU ARE THROWING OUT A RED-HERRING.

WE KNOW WHAT YOU SAID.
IT'S RIGHT HERE.

YOU SAID:

But I think the evidence is mounting that KERRY did not win the popular vote, and that the Great Exit Poll Discrepancy was largely due to poor random sampling. I realise most on this forum disagree. I can only say that I have not come to this view through bias. I was massively biased in favour of finding the opposite.

Thanks for having me. I share your pain.

Lizzie

(Elizabeth Liddle)

[Febble]

191. Well that was exactly what I meant

"I think the evidence is mounting that Kerry did not win the popular vote."

I think the jury's still out on Ohio, and if he'd won Ohio, he'd be president.

But I now think it is unlikely he won the popular vote.

What did I say that was the opposite?

{edit) Just discovered I did indeed type the opposite of what I meant. Serves me right for trying to post before coffee. Clarification in adjacent post)

[Febble]

192. Sorry, TIA you are quite right

I mistyped my response. Honestly, I got straight to it as soon as I woke up, but should have had my coffee first.

Here I what I think:

That it is possible that voter suppression and possibly fraud cost Kerry Ohio, and thus the presidency.

That it is unlikely that he won the popular vote.

Sorry again, my mistake.

[TruthIsAll]

194. Ok, now WHY is it UNLIKELY that Kerry lost the popular vote?

Where is YOUR evidence to that effect?
Are you now claiming that the Exit Polls were biased?
What is your rationale?
Is it rBr?
Or is it something else?

YOU HAVE NO EVIDENCE!

I HAVE EVIDENCE: 86 OF 88 TOUCHSCREENS TURNED KERRY VOTES TO BUSH.
I HAVE EVIDENCE: 97%+ OF ELECTION INCIDENTS FAVORING BUSH.
I HAVE EVIDENCE: SUPPRESSION OF ACCURATE RECOUNTS IN OHIO, ETC.
I HAVE EVIDENCE: MITOFSKY HAS NOT RELEASED PRECINCT LEVEL DATA.
I HAVE EVIDENCE: 22 OUT OF 22 STATES IN THE EASTERN TIME ZONE DIVERTED TO BUSH FROM THE EXIT POLLS.

I HAVE EVIDENCE: IT WAS IMPOSSIBLE FOR BUSH 2000 VOTERS TO COMPRISE 43% OF THE 2004 ELECTORATE, AS STATED IN THE FINAL EXIT POLL. THE ABSOLUTE MAXIMUM WAS 39.8%. THERE WERE 50.456 MILLION 2000 VOTERS AND APPROXIMATELY 1.77 MILLION DIED 93.5%). YOU CAN DO THE MATH. YOU'RE GOOD AT THAT, AREN'T YOU?

I HAVE EVIDENCE: MY EXIT POLL RESPONSE OPTIMIZATION MODEL CONFIRMS USCV'S SIMULATION ANALYSIS. FEASIBLE PARTISAN RESPONSE RATES CAN'T FIT THE WPE'S REPORTED BY MITOFSKY AND ALSO SATISFY THE 53% AVERAGE RESPONSE RATE CONSTRAINT. THE SAME GOES FOR THE 1.12 ALPHA.

COME ON FEBBLE, YOU ARE ON A LOSING TEAM.
LET'S PROVE rBr is FICTION.

GIVE ME SOME INPUT FOR THE OPTIMIZATION MODEL.
I WILL RUN AS MANY CASES AS YOU WISH.

WOULD YOU BE INTERESTED IN TESTING YOUR FEBBLE FUNCTION?
WELL, LET'S FEED YOUR ASSUMPTIONS INTO THE MODEL.

TAKE A LOOK AT IT HERE:
NO STANDARD DEVIATIONS, NO SIMULATIONS.
JUST PLAIN OLD CONSTRAINED OPTIMIZATION.
PLAIN OLD ALGEBRA.
YOU LIKE MODELS?
YOU'RE GOING TO LOVE THIS ONE.
I THINK.

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x375270

LET'S GET TO IT.
IT'S GONNA BE FUN.

[LightningFlash]

202. Again, Febble. You've proven time and time again...

That you at least have the ability to be fairly impartial, correct?

Then why don't you address these questions and concerns?

The issue is not whether the afroementioned model delivers fraud, it's in the fact there is not a concurrent measuring stick yet developed which CAN demonstrate fraud.

What you have here by Kathy Dopp certainly didn't do it, but it did suggest possible fraud. What you have here called the bias index didn't do it but it certainly did suggest possible bias.

What we have seen is an amount of constructs and fancy functions, but nothing concrete. And if in fact you do not wish to help Mitofsky prove a bold faced lie, I think it's time you take up TruthIsAll's challenge.

Certainly TIA's model isn't perfect as well, but it is based off a far more accurate measuring stick. He's simply taking the computations side by side the US Government's own cencus reports, crossing them against the exit polls.

Anyone can do that optimization analysis, even me.

http://www.census.gov/population/www/socdemo/voting/cps2004.html

You can do them with concrete, backed up models which do indeed demonstrate fraud. The problem is they disprove literally every instance of the RBR sampling bias, unless by some miracle, Bush received 55% or more refusers from the U.S population. That is if you're willing to move on from arguing with the government, who has put all their cards on the table.

It all comes down to the truth that yes there was fraud and Kerry won by either a small margin or a large margin, we just don't know how much. But we know where the money is don't we?

http://news.independent.co.uk/business/news/story.jsp?story=644159

Febble do a job for America instead of some scholar, I don't think Mitofsky needs your help at being dug out of the ditch.

[TruthIsAll]

216. febble, pretty please, show me the evidence...

As Adlai Stevenson said at the UN in October 1962 to Ambassador Zorin of the Soviet Union...

I'm prepared to wait until hell freezes over.

I showed you mine.
Now you show me yours.
You have already been caught with your knickers down.

Febble, you are such an intelligent person, and I know from your writings you are a good one at heart, but you have been drawn into a web of deceit. I almost feel sorry for you as you suffer the pain of acting in a way that is contrary to your nature.

Cognitive dissonance.
Forced denial of one's true belief.
So Orwellian.

Those of us who are truly independent (and knowledgeable) do not suffer from that debilitating disease.

No money in it.
Just truth.
To be able to sleep at night.
To live with yourself.

Priceless.

[Febble]

225. Thanks for the nice comments, TIA

(though I could have done without the bit about the knickers - leave a menopausal woman a bit of dignity ;-).

And I do understand, that if you do believe that I am, as I believe you to be, essentially good and intelligent, why my stance must seem incomprehensible to you.

I think the evidence that the 2004 was corrupt is compelling, both as regards fraud and voter suppression.

I think it is possible that the corruption contributed to the exit poll discrepancy. Where we differ - and it is a huge difference - is in whether we believe that the magnitude of that discrepancy is commensurate with the scale of the fraud.

The reason I wrote, the in post you quoted earlier, that the evidence against massive fraud was mounting, was because of some specific pieces of evidence, neither of which you are likely to accept (I know you do not) but which I, for reasons which probably stem simply from my background in behavioural science, found at that time persuasive.

These were simply the evidence regarding the WPE data in the E-M report. It was an inadequate report in many ways, some of the ineadequacies being ones you yourself have cited: that hypotheses regarding fraud seem to have been given cursory attention; that no statistical details were given of the tests used; that no multiple regession results were reported; and that, IMO, the dependent variable, the WPE, is a flawed measure.

However, I do not agree that what was presented was "no evidence" for the differential response hypothesis. Sampling bias is a survey design problem, and will tend to be greater where sampling is non-random. The report clearly showed that the the "red shift" was indeed greater where sampling protocol was most likely to be compromised. As a statistician, this is strongly suggestive that sampling bias accounted for at least part of the "red shift", although it does not in itself explain why the shift itself should have been red. It is nonetheless a well-known phenomenon in behavioural research that where an underlying response bias is present, that it will be ampligied by poor random sampling protocols.

However, a further piece of evidence that for me, heaped the scale on the side of response bias, rather than fraud, as the likely more prevalent contributor the the "red shift" was the lack of a linear relationship between vote-count margin and bias. At first, and at the time when I wrote my paper, I simply regarded the postulated (at that time) lack of linear correlation as evidence against the hypothesis that fraud was concentrated in "Bush strongholds", and considered that all it would show (and in the event did show) was that the problem, whether fraud or sampling bias, was randomly distributed across the whole range of precincts. However, my own modelling exercises have since convinced me that any massive fraud, i.e. fraud on a scale large enough to account for even a substantial part of the "red shift", would produce a linear slope between bias and vote-count margin, essentially because of the Gaussian distribution of the precincts. Fraud would shift the precincts right-wards, into Bush territory, and upwards, into bias territory. This would tend to induce a slope in the correlation that is absent in the data. However, one remaining hypothesis, to my mind, that might account for massive fraud, but nonetheless be consistent with the observed non-significant slope, is that fraud may have been concentrated, not in Bush strongholds, but in Kerry strongholds - thus counteracting the slope. But at this point, I think the more parsimonious explanation (and Occam's razor applies here) is simply that fraud, if it occurred, was not on a scale to tip the slope. I may be wrong. I am simply saying that that was why I said the evidence is mounting.

I don't think the exit polls will ever provide conclusive evidence that either fraud played no part in the discrepancy, or that it did. I think they are simply too crude a tool for auditing elections. It is why I so strongly support your campaign, and the campaign of USCV to fight for fair, secure, and auditable elections. But I do think the exit poll data needs to be subjected to re-analysis, and I hope this will happen. The E-M report was simply not convincing.

Anyway, I wish you well. I am not denying my innermost beliefs when i say these things. I am simply trying to be as honest as I can. No, I don't sleep well at night. I take what you, and others on DU say both about me and about my arguments, seriously. But I cannot say what I believe to be either false, or unsupported. And at present, my conclusions regarding the exit polls simply differ from yours.

Peace? :)

(edited after posting as I clicked post instead of preview)

[autorank]

226. Enough of the "era of good feelings"... back to work.

Good morning. now where were you and TIA

:hi:

[autorank]

140. Well, why don't you "debate the numbers"' and leave if "abused"

So far the Mitofsky, MP, whatever faction seems to be interested in intra-mural academic debate (Tfc's term) rather than helping in the process of refining TIA's model. It was a friendly offer by TIA. Why not give it a try?

[OnTheOtherHand]

91. "internal squabbles"? discourse? democracy?

Look, if folks here can wrap their heads around the idea that it is possible to suspect (or even believe) that Bush won the 2004 popular vote -- and still think that potential election fraud is a huge problem -- then not only will we waste less time throwing pots at each other, but we will have better politics. Because there is pretty much no possible way to convince most people that Bush stole the election (many won't even believe that he reneged on his ANG obligations or willfully misled the country into war), but lots of people are creeped out by "just trusting computers."

Academics spend a lot of time worrying about evidence, for obvious reasons. Bruce O'Dell is an engineer, and likewise spends a lot of time worrying about evidence, because his steadfast conviction that something ought to work is not in itself a warrant that it will work. Lots of people here who are neither academics nor engineers spend a lot of time worrying about evidence. If we can't even argue respectfully about evidence, then we might as well all make stuff up. And maybe we should consider renaming the board the Dogmatic Underground. If assuming that the evidence proves fraud in 2004 is the price of admission to the discussion, then there isn't much to discuss.

[LightningFlash]

92. Another strawman in a hat does not make a sole.

You're incorrect to cite that using other accredited non-statistician PHDs to forward an argument they don't understand is advantageous to the process. Bruce is an engineer, and has never claimed to be a mathematics professor.

There will never be one SOUND decisive measuring tool of fraud, only logical guesses. But when it comes back to the government's own decreed cencus information, that provides about the only measuring tool that gives credence.

Using a ruler, a model or any such made up tool to measure improbable results does not a proper truth-seeker make. The strawman argument has been destroyed, and it will not be brought back.

The US Cencus gives the full and complete picture, the only indicator we should dare rely on since Voter News Service refuses to release their data. They and Mitofsky continue to hold that position, and firmly grasp the reality they are enabling the suspicion and discourse themselves. Let them have it, debate over.

[OnTheOtherHand]

94. I'm sorry, but you are wrong about the census data

(Neither Bruce O'Dell nor Kathy Dopp has a Ph.D. But never mind.)

As I pointed out on another thread, the Census Bureau study you presumably refer to is not "the US Census," and it doesn't even claim to give a full and complete picture. The estimates are drawn from the CPS (Current Population Survey).

Here's a Census Bureau press release about the 2000 election (copied from my other post) --

http://www.census.gov/Press-Release/www/releases/archives/voting/000505.html
or http://tinyurl.com/4k6th

It ends: "Data are from the November 2000 CPS. Statistics from surveys are subject to sampling and nonsampling error. The CPS routinely overestimates voter turnout. As discussed in greater detail in the report, the CPS' estimate of overall turnout (111 million) differs from the 'official' turnout (105.6 million votes cast), as reported by the Clerk of the House."

Notice the word "routinely." It wasn't just in 2004, it wasn't just in 2000.

[LightningFlash]

113. I'm surprised this even is contested. Do wonders never cease?

The U.S. F.B.I or some "secret" order doesn't make the Cencus results. The official, confinaggled United States Cencus Bureau makes the results. They have done so by compiling Cencus data for the U.S.A the last 15 years. They probably are more of an expert than some foreign Book Writer, an Engineer, a hundred other people who are not renowned and accredited mathmeticians. That says it all.

They are accurate and in totality, and also state under every certain term that the error rare is.30%, and I'm sure they would be the LAST people to put forward a strange hypothesis like everyone lies to the cencus takers.

Their job is to report the facts instead of cover it up, what you are proposing instead is a conspiracy that would allow "vote fraud theorists" to be duped and then dan-rathered. It's even more out of the question than thinking this election was stolen.

What it provides and clearly demonstrates, is a heavy amount of fraud and areas of sampling bias.

http://www.census.gov/Press-Release/www/releases/archives/voting/004986.html

"Note how close the given MoE is to the calculated MoE (within
0.02%)using the formula:
MoE= 1/sqrt(n), where n= population size (000)

Census MoE
U.S. Total Pct Pct MoE (000) 1/sqrt(n) Diff
.Total 125,736 58.3 100% 0.30% 377 0.28% 0.02%

.Male 5
8,455 56.3 46.49% 0.40% 234 0.41% -0.01%
.Female
67,281 60.1 53.51% 0.40% 269 0.39% 0.01%

.White alone
106,588 60.3 84.77% 0.30% 320 0.31% -0.01%

..White non-Hispanic alone
99,567 65.8 79.19% 0.30% 299 0.32% -0.02%

.Black alone
14,016 56.3 11.15% 1.10% 154 0.84% 0.26%
.Asian alone
2,768 29.8 2.20% 1.70% 47 1.90% -0.20%
.Hispanic (of any race)
7,587 28.0 6.03% 1.20% 91 1.15% 0.05%

.White alone or in combination
107,930 60.3 85.84% 0.30% 324 0.30% 0.00%
..White non-Hispanic alone or in combination
100,726 65.7 80.11% 0.30% 302 0.32% -0.02%
.Black alone or in combination
14,324 56.1 11.39% 1.10% 158 0.84% 0.26%
.Asian alone or in combination
2,980 30.7 2.37% 1.70% 51 1.83% -0.13%

Table 4a. Reported Voting and Registration of the Total
Voting-Age Population, by Sex, Race and Hispanic Origin, for
States: November 2004"

[OnTheOtherHand]

128. I can't tell whether you even realize that the Census

is conducted every 10 years, while the CPS is "a monthly survey of about 50,000 households conducted by the Bureau of the Census for the Bureau of Labor Statstics ."

http://www.bls.census.gov/cps/bovrvw1.htm

And I certainly can't tell why you think (if I understand rightly) that the CPS estimate for 2004 was more accurate than the CPS estimate for 2000, when the exit poll result was much closer to the official returns.

[LightningFlash]

130. Virtually incredible.....Now it's the type of Cencus that matters!

I can see you are really far reaching now......Perhaps you do not yet realize, there is no other alternative now but to accept that fraud tilted the election?

:wow:

You dispute the very merit of the US Bureau of Statistics.....Fascinating..

[OnTheOtherHand]

159. what, are you kidding?

For one thing, there is no US Bureau of Statistics. There is a Bureau of Labor Statistics, and it is very good, as is the Census Bureau. Of course that doesn't mean that their numbers are infallible; one of the things that makes them very good is that they know a lot about the problems with their own numbers.

The Current Population Survey is not a "type" of census in ordinary English. A census is a population enumeration, an attempt to count every person in the United States. The CPS is a sample survey.

The Census Bureau itself, in a passage I quoted and linked to, stated, "The CPS routinely overestimates voter turnout." So if you insist on using the CPS as an accurate measure of voter turnout, who is it that is questioning the merit of government statisticians?

[autorank]

98. OTOH

""not only will we waste less time throwing pots at each other" --The O'Dell memo is a big pot thrown at Ron Baiman. Give me a break.

[OnTheOtherHand]

127. No, it's a big pot thrown at his work. We distinguish. n/t

[autorank]

203. Go here if you realy want to contribute to election fraud research

http://upload.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x375366

[Time for change]

95. I have several points to make about this

1) The two main participants in this debate are both members of US Count Votes (USCV), a voluntary organization that is dedicated to analyzing data to determine what happened in the 2004 election AND advocating election reform. They are both dedicated to this very important effort, they have been working very hard at it for months, and I doubt that they have received any money for these efforts. So I think that we can learn a lot by listening to what they both have to say. Yes, Bruce O'Dell's statement was incendiary. That's what happens sometimes when people debate emotionally laden issues. But at least they refrained from personal attacks or questioning each other's motives.

2) In the hope of furthering this debate along, I recently posted a thread entitled "Does Mitofsky's AAPOR presentation support his conclusions": http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=373817. In the first sentence of this post I say "No, I don't think so", and then I go on to explain why I believe that not only did the information presented by Mitofsky at AAPOR not support his rBr hypothesis, but it may argue against it.

Unfortunately, IMHO, that thread didn't receive a lot of attention, and as soon as I posted this thread it didn't receive any attention at all (I think that a large part of the reason for this is that, unlike my previous post that you refer to, this one used a statistical language that most DUers don't understand because they don't work with statistics regularly.) So, if you want this debate to be balanced out by somewhat of a counter argument (though nobody, including Bruce O'Dell, is arguing for rBr), then why don't you go back and start the discussion going again on that thread?


3) In the thread that I refer to above, one of the arguments that I mention refers to the USCV report that concluded, through the use of mathematical simulations, that analysis of Mitofsky's own data leads to implausibly high Kerry voter completion rates in the Bush stronghold precincts. As you know, this report has come under a lot of criticism, especially from USCV's own Vice President, Bruce O'Dell.

I honestly do not understand this argument well enough to make useful comments on it. However, since the USCV report was under attack for this I found another way, without using mathematical simulations, to calculate what the Kerry and Bush voter completion rates would have to be in the 40 Bush stronghold precincts combined. What I came up with (assuming a Bush vote count for those 40 precincts of 85%) is 75% Kerry voter completion rate, versus 53% Bush voter completion rate -- something similar to what the USCV report came up with. I don't know if that bolsters the USCV report or not. But I do feel, though other honest people may and do disagree, that those numbers seem at least somewhat implausible in the absence of fraud. Maybe I should start a new post on that.


4) My posting of Bruce O'Dell's statement on the DU did not make this issue public, because it was already public. What it did do, in my opinion, was provide interested DUers with the opportunity to become more familiar with the issues and engage in a debate which could lead to a better understanding of the issues and therefore a greater ability to discuss them in public.


5) You say that I mentioned that I'm a professor, and asked if at my university we engage in incendiary debate like this. I have never said on the DU that I'm a professor. I am not currently a professor, but I used to be an Assistant Professor of Preventive Medicine at the Medical College of Wisconsin.

Currently I am an epidemiologist who works for the FDA. It's interesting that you made the comment about public incendiary debate. Less than a year ago I attempted to publish an article in the Journal of Vascular Surgery that questioned the safety of a medical device that I had been studying. Most people might not have considered the article to be incendiary, but the manufacturer of the device certainly did. The Journal accepted the manuscript for publication, and it was about to be published, when suddenly the manufacturer found out about it, complained about it to the FDA, and the FDA ordered the Journal to pull it. The Journal pulled the article, as requested by the FDA, though the editor of the Journal wrote an editorial in his own journal, which was highly critical of the FDA for doing this. And then somebody leaked the whole story to the Wall Street Journal, which published it on their front page. So I'm used to incendiary arguments being aired in public, and I can tell you that I don't regret for a minute that this issue was aired in public, or the role that I played in it. Sometimes airing things in public helps greatly, by helping people to become better eductated about things.

[autorank]

100. Thanks for sharing. This whole thread is a distraction, IMHO

PARALYSIS BY ANALYSIS...BY DESIGN, PERHAPS

[LightningFlash]

90. Time for Change, it's time to live up to your name.

What you're presenting here is admirable, but it's most definitely in the wrong context. This debate on what is an actual measurer of applicable fraud is one of the worst diversions that the right-wing run Mitofsky ring could possibly ever come up with, and true to form it has already been completely debunked by the actual cencus results.

We don't "prescribe" any one's personal measuring tool as the official source. Tell that to Bruce, highly and concretely, since he is not a statistician by his own admission. What we most certainly prescribe, is using the US Government's OWN cencus data which has already been accurately measured, and used for over 15 years.

http://www.cencus.gov/Press-Releease/www/releases/archives/voting/004986.html

And we can assuredly trust their own facts which confirm single handedly, response bias was not varied enough, nor even possible in the 2004 election to create the anomoly. Vote "irregularities" and fraud most certainly was, however. Using our own cencus data as a model regression can show just how tentative it was. Debate over.

[Time for change]

97. Sorry Lightning Flash, but I can't pull up your link

Therefore, I don't understand the point you're making, and I can't respond. Perhaps if you could explain within the post what point you're trying to make, I could say something about it.

But I do want to make one thing clear: I have never in the past, nor do I now support Mitofsky's rBr hypothesis (nor does Bruce O'Dell BTW).

As a matter of fact, as soon as the report came out, I believe within 24 hours of its release, and well before the USCV report, I posted a thread on the DU criticizing it:
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=295225

Then recently I posted a thread that attempted to explain to non-statisticians the essential issues of the controversy:
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=371726

And even more recently, I posted a thread that attempted to explain why I believed that even Mitofsky's AAPOR presentation did not support his hypothesis:
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=373817

So there is no need to convince me of the weakness of this hypothesis. However, where I stop short is in saying that the case against rBr has been proven. I don't believe that it has been. But I do think that if we keep studying it, pressing for the release of more data, and engaging in productive debate we will take steps towards a better understanding of this issue.

[autorank]

103. More of the same...thanks for sharing...very helpful (yawn)

[kster]

120. Let me know when

something good happens


:boring: :boring: :boring: :boring: :boring: :boring: :boring:

:boring: :boring: :boring: :boring: :boring: :boring: :boring:

[autorank]

121. A hearty LOL generated by the text

:hi:

[Name removed]

221. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

222. Deleted message

Message removed by moderator. Click here to review the message board rules.

[LightningFlash]

260. Turning around in circles, ignoring the real problem: The Phantom.

Bush would have needed 55.67% of united states refusers, in order to even complete the data or come near the exit poll discrepancy. He would have needed it in only eastern states.

With the final facts being the last, adjusted weighting has proven bogus how can you possibly assume that there is virtually any validity to the hypothesis?

The gender weightings reversed, with the female vote becoming smaller

1.5 million of those bush voters who voted in 2000, were included, even though they had died(at minimum)

The switching of weighting on the religious chard ID can not be explained (as per http://exitpollz.net )

You are no closer to getting underneath any of these bogus demographics, than you are to saying the exit polls were useless.

The exit polls in Florida state have, just the opposite, proved fraud with the analysis completely backing it up.

""Miami-Dade County, Florida. May, 2005. ES&S iVotronic.
New evidence shows both phantom votes and lost votes in the November election.
The number of voters reported by election workers didn't match the number of ballots cast in 260 (35%) of Miami-Dade's 749 polling places. Some showed more votes than voters ("phantom votes"); others showed significantly more voters than ballots cast.*

While some of the discrepancies can be traced to sloppy procedures and training, others are evidence of problems not yet explained: "


Phantom votes: precinct with a major difference was Precinct 362, which recorded 583 votes and 859 signatures. A review of that precinct's signature log found 580 signatures.
Phantom votes: in Precinct 41 there were 910 votes and 844 signatures, a difference of 62.

Lost votes: At combined Precinct 117/166, the ballots totaled 995 and the signatures numbered 1,276.

Phantom votes ... then more phantom votes: For Precinct 816, in the Church of the Ascension at 11201 SW 160th St., a Review inspection of the voter log showed 945 signatures, while the iVotronic computer tape showed a count of 1,032. But the individual machine counts are listed on the tape as well, and they add up to 945. But the number certified by the canvassing board came in at 1,116 votes.

In this case, the discrepancy may have been due to an iVotronic machine malfunction. Lynn Kaplan, a volunteer observer for the reform coalition who was at that polling place on Nov. 2, said in an interview that as a poll worker was closing down one of the iVotronic machines at the end of the day, an error message popped up on the machine's digital screen saying: "Internal malfunction/unit closed to save data/vote data corrupted."

... said the situation in Precinct 816 highlights the need for better procedures and training for reconciling signature and ballot totals. She also stressed that the county must investigate discrepancies to see if they resulted from fraud or equipment malfunction.


The Secretary of State's office has refused to get involved.

Jenny Nash, a spokeswoman for Florida Secretary of State Glenda Hood, said the issue of reconciling signature totals and machine counts was a local one and is not the concern of her office. "Each supervisor has their own process for how they reconcile the numbers," Nash said.

* Elections Discrepancies found in 35 percent of Miami-Dade precincts. Daily Business Review. May 06, 2005 By: Jessica M. Walker;"

http://www.dailybusinessreview.com/news.html?news_id=34733 "
http://www.votersunite.org/info/content/newmessup-17.asp

http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01008-machine7301.PDF
http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01006-machine6545.PDF
http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01004-machine21285-oct.PDF
http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01004-machine21284.PDF

Proven wrong. Over and over again. And not one single point is responded to!

"How can the discrepancy actually be that large?"

Not unless the exit polls had a miracle, and trained everyone to do computer-laser accurate oversampling of voters everywhere and did it so well it moved mountains.

Or, the valid and final explanation: Machine manipulation. Which can and DID vary the discrepancy completely in the state of Florida, where there was only about 10% uniform random bias...

[sunshinekathy]

244. Kathy Dopp here - Even Mitofsky doesn't agree with Bruce O'Dell

I'm sorry to join this conversation so late, but I've been on the road and Bruce posted a bunch of anti-USCV posts before I even had a chance to get home and begin to edit and add to our current working paper.

Just some brief points:

1. Even Mitofsky is now saying that he "never said that the exit poll discrepancies could be explained by a response bias with a constant mean". Unlike O'Dell, Mitofsky understands that the math as demonstrated by Ron Baiman and myself in our recent papers clearly refutes the possibility of a response bias with constant mean.

2. O'Dell apparently used highly implausible LARGE variances to get his erroneous conclusion. The variances he assumes would make any glob of data have a constant mean, and are not justified.

3. Bruce does not seem to understand the difference between saying that possible explanations of the exit poll discrepancies "include" certain possibilities and "require" these same possibilities. So Bruce continues to misrepresent and distort what our current working version of our paper says as well as misunderstand its mathematics, despite that I have pointed out many times to him via email that he is distorting and not honestly reporting the contents of our current working paper. He has not located a specific flaw with any sentence, mathematical or logical assertion in our current working paper, despite my repeatedly asking him to so that we could discuss it to his satisfaction.

4. Bruce's own numbers can probably be used to refute his own conclusions, but we are very busy and haven't had time to spend on it. There are a gazillion important projects in the works that must be done. USCV REALLY needs funding and more staff. I have set aside a bedroom w/ private bath in my home with its own entry and porch for someone who would like to come work and live in Park City Utah and help operate the business of USCV and be provided room and board.


5. I have informally asked Bruce O'Dell to either make a public retraction or resign from the Board of USCV and to resign as Vice President due to his destructive actions against USCV both on our statisticians list and now publicly.

6. Not even ONE PhD mathematician or statistician in America has refuted anything in our current working paper or in our prior paper to my knowledge (please inform me if I'm wrong), with the exception of the one minor mistake we found ourselves in our last finished paper that we are going to clearly correct in our current working paper prior to its finalization. We did not catch it until we did the analysis of the patterns of exit poll discrepancies that are produced by vote shifts.

I do not know why Bruce has done the destructive things that he has done ever since the AAPOR conference. Bruce O'Dell donated a lot of his own money and time and was one of the founding members of USCV and contributed a lot to USCV. We are very disappointed that he has become so destructive of USCV in recent weeks and we are hoping he will come around and have tried and are still trying to get him to recognize and publicly admit his mistakes, but it doesn't seem likely considering how long now he has stuck to them, and his inappropriate behavior. I am surprised that someone who is usually so rational has stuck for so long to a mistake.

Bruce is 100% wrong on his mathematical claims, as Mitofsky realizes, even if Bruce, and perhaps Liddle, do not.

It is unfortunate however, that Bruce is muddying the waters so, as most people do not understand the mathematics enough to know who is right and who is wrong.

Best,

Kathy Dopp
http://electionarchive.org

[tommcintyre]

245. Thanks for taking the time to come here and explain all this Kathy n/t

[OnTheOtherHand]

246. waist deep in the Big Muddy

I do not address Kathy Dopp directly, for reasons she knows. Just some mostly brief responses:

1. As close observers of the debate know, it is almost impossible to tell what any statement about "constant means" actually means out of context. Some of us have tried to engage this issue in detail on DU so that people can form their own conclusions about what is really at stake. There is room for honest disagreement. I see no room for the inference that Mitofsky disagrees with O'Dell. Both agree -- I think everyone agrees -- that the observed mean alpha is higher in the 40 high-Bush precincts. The question is, So what?

2. The mean absolute WPEs reported by Edison/Mitofsky in January are (high-Kerry to high-Bush) 8.8, 13.4, 15.2, 13.2, 12.4. The mean absolute WPEs I calculate from Bruce O'Dell's simulation are 7.7, 11.9, 13.1, 11.8, and -- take your pick -- 12.2, 10.6, or 14.2 for three simulations at the high-Bush end, all of which give similar overall results. I see no basis for Dopp's confidence that his variances are "highly implausible LARGE." The scatterplots make clear that the variances are large.

3. Several DU participants are finely attuned to "PR" and "framing," so I invite them to explain to me what the USCV working paper actually says or proves. I'm not sure, and I actually wrote part of it. But if anyone is convinced that it refutes Mitofsky, you will have to explain that as well. Meanwhile, Dopp is cutting the distinctions pretty fine here.

4. O'Dell wrote a 33-page paper. Dopp writes this. A legitimate choice. But it obviously doesn't support her triumphant conclusions that O'Dell is "100% wrong." Good luck with that.

5. I think it takes some chops to insist on releasing a working paper despite vigorous internal dissent, and then to complain -- well, more than complain, really -- when people aren't willing to shut up and toe (or mouth) the line.

6. Is someone about to denounce Dopp for her elitism? The invocation of PhDs is a red herring -- neither Dopp nor O'Dell has a PhD, and who cares, anyway? It's open to interpretation, I guess, whether anything in either the USCV working paper or O'Dell's response has been "refuted."

Folks, USCV is not a Maoist cult. There is no reason whatsoever why the vice president should have to recant the thoughtcrime of having dared to disagree with the president. Dopp chose this fight, not O'Dell. I think it was a lousy choice.

[sunshinekathy]

248. Yes, O'Dell should recant the false statements & incorrect conclusions

There is every reason in my opinion, that everyone should recant any false statement or mischaracterization they make. Bruce has made false statements and mischaracterizations about our working paper.

I would not hesitate to recant something that I've said that was incorrect in a second, as I did the first minute I woke up this morning, without being asked to, and without having to have my error pointed out to me, about my exageration about Bruce's math being "100% wrong". (When I should be driving down the road, I made a special trip back into the restaurant where I can pick up Internet access to recant. I expect the same honesty and intellectual integrity from Bruce O'Dell.)

It is also my practice to immediately recant any false conclusions I have made as soon as I find out that they are incorrect as we have done in our current working paper about one of our statements in the prior paper, and will do so more forcefully so that everyone understands it better prior to finalizing our current working paper.

Bruce could also recognize that his conclusions are incorrect and recant those also. Mitofsky has recently said that he never said that a constant mean distribution of response bias could explain the exit poll discrepancies. I believe that Mitofsky, Baiman and I have a much better feel for the implications of the mathematical analysis and that O'Dell's conclusions are simply incorrect.

That's all.

I have had such an intuitive feel for mathematics ever since I was young that I do have trouble seeing how O'Dell could be making the conclusions that he has, I admit. Even my son passed his calculus AP test when only a sophmore in highschool. It runs in my family. Plus I am over-whelmed with productive tasks that need doing.

We have many opportunities to set things right again in America that must not be missed and we need help getting those tasks done in time.

Kathy

[OnTheOtherHand]

249. I daresay he will retract them as soon as someone refutes them

Dopp's intuitive feel for mathematics has not prevented her from flaming me (and, I'm sure, others) for supposed egregious errors, only to admit later that she was wrong. It has not prevented her from asserting that parts of Febble's work never made any sense, only to admit later that they were correct. So I can vouch that she is willing to admit errors. I can also vouch for Bruce O'Dell in that regard. But if anyone thinks that Dopp's intuitive feel simply settles the debate -- well, that isn't thinking.

I do not give Dopp huge points for characterizing her retraction of the ridiculous "100% wrong" business as a "slight correction," even if she had to make a special trip to the restaurant to do so. Your Mileage May Vary.

[autorank]

250. And you'd been doing so good...

Contact the DNC and Give 'em Hell About NOT Acting on Election Fraud

NEW LEADERS FOR A NEW DEMOCRATIC PARTY

[tommcintyre]

258. Well, looks like I need my waders AND shovel. THERE YOU GO AGAIN ;)

Well, looks like I need my waders AND shovel. THERE YOU GO AGAIN

<You said:>
"Some of us have tried to engage this issue in detail on DU so that people can form their own conclusions..."

Really? It seems to me, the main intent of the "Some of us" you refer to, has been to spin and obfuscate. For example, in this thread:
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=371175#371617
"A quick summary of what I've seen here..."
"Basically, what I saw was a pattern of CLEAR questions being directed to the "implausibility of rBr" doubters (to explain their pro-rBr claims); and vague (obfuscated answers).”

Yes, I think it IS a good idea that: "people can form their own conclusions". How about if they go read the exchanges between you and eomer in this thread and do exactly that - to see WHO is spinning (obfuscating), and who is seeking clarification?

BTW (from the same thread): Do you still think USCV is involved in a cover-up? I don't think I ever got a direct (clear) answer:
"So which IS it really? USCV cover-up, or not?"
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=371175#371993

<You said:>
"...it takes some chops to insist on releasing a working paper despite vigorous internal dissent..."

Does it really? So a "rogue" (ODell) tries to interfere with the progress being made in fine-tuning the exit-poll-discrepancy-points-to-fraud argument, and you "spin" it that they shouldn't have released the new information?

<You said:>
"...more than complain, really -- when people aren't willing to shut up and toe (or mouth) the line."

Yes, it is DEFINITELY "more than complain, really". HIS RESIGNATION IS BEING ASKED FOR (and understandably so - his actions/words speak for themselves). In times not-too-far in the past, people in situations like this would've had the integrity (or whatever) to step down. (This kind of reminds me when Michael Eisner refused to resign, even though over 40% of the Disney stock-holders gave him a vote of no confidence.)

<You said:>
"Is someone about to denounce Dopp for her elitism?"

Didn't you just do that in a back-handed way with your question? BTW: I don't see anyone else doing it, do you? (Maybe you can get some of the "Some of us" you mention above to join in - so YOU are not so starkly alone? ;) )

Also, the way you used "elitism" struck a chord. "I've seen this before", I thought. Doesn't the Bush admin use this as a "hot button word" (like liberal), to get people not to think, but, instead, react emotionally? So I did a quick Google search, and there it was:

http://www.deltos.com/reference/SJR/archives/000068.html
"Anti-Elitism in the Bush Camp

I caught an article in Washington Monthly that seems to hit a nerve. Titled Creative Class War: How the GOP's anti-elitism could ruin America's economy, it's an interesting read. I'm appalled by the bubba-platform ("don't vote for smart people!") that the GOP runs (and wins) on."

<You said:>
"The invocation of PhDs is a red herring -- neither Dopp nor O'Dell has a PhD, and who cares, anyway?"

The qualifications of the signatories in such a highly technical field as statistics is DEFINITELY NOT a "red herring" (a distraction). It IS (and should be) a KEY CONSIDERATION in judging the validity of differing positions. This is true since the vast majority of people do NOT have a background in this very complex field. Consider the following example to illustrate this point: If you had a serious medical condition, whose advice would you take more seriously; a fully trained and licensed medical doctor, or someone who just had an interest in, and used health information in their job (for example, like someone who works in a health food store)?

So, when it comes to a statistical matter as serious as whether the election was fraudulent, shouldn't we listen more to a fully trained and licensed (PhD) statician - not a computer specialist (ODell), a psychology and music specialist (Febble), or a political science teacher (OTOH)?

So Kathy is VERY CORRECT (she is, herself, a mathematician). So, it is VERY SIGNIFICANT when she says: Not even ONE PhD mathematician or statistician in America has refuted anything in our current working paper....

---------------------------------------
Finally,

<You said:>
"...USCV is not a Maoist cult."
<No one has said they are – or even implied they are (except you?). ARE YOU implying they are, if they don't agree with you?>

<You said:>
"...should have to recant the thoughtcrime of having dared to disagree with the president."
WOW! "thoughtcrime"! That's straight from Orwell's "1984"! (See link below.) Orwell wrote about a fictional fascist society where "Big Brother" even tried to control people's thinking. So are you trying to manipulate us into thinking USCV is fascist if they don't agree with YOU on this issue?
http://www.newspeakdictionary.com/ns-dict.html#crimethink

And, to be accurate, Kathy was CLEARLY referring to ODell's ACTIONS, not thoughts.

Examples from Kathy's post (bolding mine): 1) "Bruce posted a bunch of anti-USCV posts..."; 2) "...he is distorting and not honestly reporting the contents of our current working paper."; 3) "...asked Bruce O'Dell to either make a public retraction or resign from the Board of USCV and to resign as Vice President due to his destructive actions against USCV..."; 4) "I do not know why Bruce has done the destructive things that he has done ever since the AAPOR conference."; 5) "We are very disappointed that he has become so destructive of USCV in recent weeks..."

So, I "invite" YOU to explain exactly WHERE USCV has criticized ODell for his thoughts, rather than his destructive actions, that are CLEARLY counterproductive to the USCV efforts?

<And, finally, you say:>
"Dopp chose this fight, not O'Dell. I think it was a lousy choice."
<My, my... that's a pretty presumptive spin, isn't it? In reality, ODell "chose this fight" when he began engaging in his "anti-USCV actions" (as described above), didn't he? In contrast, Kathy is merely defending USCV from Bruce's anti-USCV actions, isn't she?

So, after-all, ODell really is the one who made the "lousy choice", isn't he?

And, frankly, I think YOU made "a lousy choice", in how you framed your post above.

[OnTheOtherHand]

266. I don't have time to read all this

There doesn't seem to be much on 2004, quite a bit on what you consider terrible things that I said here or there. Certainly no effort to refute any of the substantive points I've made.

(I do pause to wonder: who are you characterizing as a fully trained and licensed statistician? What does that mean? I believe Baiman has a PhD in economics; I have a PhD in political science; neither of us has to kneel to the other, nor should Dopp, O'Dell, or anyone else kneel to us.)

But here's a strange note you strike, that you seem to think is telling, since you put it in bold (of course, you do that a lot):

"explain exactly WHERE USCV has criticized ODell for his thoughts, rather than his destructive actions...."

That is the issue in a nutshell. Criticizing ideas is good. O'Dell criticized ideas. This is being characterized as a destructive action. That's outrageous, so I am angry.

Tom, put down the -- whatever you're waving around and try to ask yourself, "Wait, why am I trying to equate informed dissent with treason? Isn't that someone else's game?"

[tommcintyre]

274. You're NOT a statistician, Ron is -stop obfuscating PLEASE!

<The first citation that comes up if you Google "Ron Baiman"
1 The United States of Ukraine?: Exit Polls Leave Little Doubt ...
File Format: PDF/Adobe Acrobat - View as HTML
Ron Baiman, Ph.D. December 17, 2004. Economist/Statistician <my bolding so you don't miss it OTOH ;) > Ron Baiman works as a senior research specialist at the Institute ...
www.freepress.org/images/departments/997.pdf - Similar pages

<You:>
Political Studies - Assistant Professor
Research Interests:
anti-poverty policy
environmental politics
U.S. foreign and military policy
U.S. religious politics, prejudice and tolerance
public deliberation on policy issues
American politics, particularly: public opinion and "American values"
nonviolent social change movements
democratic responsiveness

Do I see "statistician", or even mathematician in there? Noooooooooooo! (I did an extensive search, and you certainly don't appear to be a statistician, or even mathematician.)

So, I guess I'll stick with Ron as a first source for statistical advice.

Now YOU... I would come to first for info on: anti-poverty policy, environmental politics, etc. ;)

The rest of your message, I won't respond to, just as YOU were too busy to respond to mine - pretty convenient (i.e. evasive, obfuscating), isn't it?

Maybe you should add those to your list of interests above too? ;)

[OnTheOtherHand]

276. gee, since we are wandering through credentials

I won't check Ron's, except to point out that "/statistician" in his description in the Free Press doesn't make him a "licensed statistician." I'm not faulting Ron's credentials, nor ascribing your strange argument to him.

I apologize if you find my description of research interests self-effacing. Public opinion analysis is indeed a heavily quantitative field, and since much public opinion research is survey-based, my training is pretty directly on point. I taught quantitative methods at Columbia's School of International and Public Affairs (me and heaven knows how many other people).

Tom, let's be honest. If someone posted a message asserting that TruthIsAll should stand down from his arguments because OTOH, who is a licensed political scientist specializing in American politics, disagrees with him, would you post to say that that was a good argument? and would it be?

Tom, I will never question your credentials to challenge my arguments. As for your willingness and ability, we will see.

[tommcintyre]

279. You just keep trying to "dance around this", don't you?

"6. Not even ONE PhD mathematician or statistician in America has refuted anything in our current working paper..."

Bold 'nuff for ya? ;)

[OnTheOtherHand]

281. ah, but how many have even bothered to read it?

AFAIK, Baiman is no more (or less) a "PhD mathematician or statistician" than I am. Why does this matter?

But hey, there may be an opportunity here. If I can get an actual mathematics PhD to criticize the USCV paper, will you give up? I am guessing no.

Speaking of dancing, you didn't answer my questions. If someone posted a message asserting that TruthIsAll should stand down from his arguments because OTOH, who is a licensed political scientist specializing in American politics, disagrees with him, would you post to say that that was a good argument? and would it be? "Not even ONE PhD political scientist in America has refuted anything I have said in this thread." I'm embarrassed to make that argument even facetiously.

[tommcintyre]

287. So maybe Mitofsky will "help" you out? ;)

http://en.wikipedia.org/wiki/Wikipedia:Reliable_sources
"Beware false authority

Would you trust a plumber to reliably and painlessly fill your cavities? We hope not. Likewise, you should probably not trust someone who has a Ph.D. in plant biology <or "Political Studies"> to tell you about quantum mechanics <or statistics>.

Just as actors in TV commercials don white lab coats to make viewers think they are serious scientists, people with graduate degrees in one field will flaunt it in an attempt to fool people into believing they are experts in a completely different field, or in all fields. Watch out for this and other false claims of authority."

<You say:>
"AFAIK <As Far As I Know>, Baiman is no more (or less) a "PhD mathematician or statistician" than I am." Why does this matter?

As Far As <YOU> Know??? WOW! now THAT'S reliable information! :sarcasm: I think I'll trust that citation above (and my previous explanations on this) before I go with your biased "guess".

Why does this matter? Well... that's pretty obvious for the unbiased; but for you - read the wiki quote above.

"If I can get an actual mathematics PhD to criticize the USCV paper, will you give up? I am guessing no."
Oh sure, I can just see it. You call up "Uncle Warren" (Motofsky), and say: "Boss, I need one of your other lackeys <with a stats PhD> to discredit this here paper". ;)

Actually, it's pretty amazing, considering ALL the publicity (from Pastor, etc.) that none have done this. Kathy is right. It speaks very strongly of the paper's soundness/validity.

<On the subject of Mitofsky, I have some links for you to ck out at the end of this post.>

"Speaking of dancing, you didn't answer my questions. If someone posted a message asserting that TruthIsAll should stand down from his arguments because OTOH, who is a licensed political scientist specializing in American politics <you mean a "Political Studies" assistant professor?>, disagrees with him, would you post to say that that was a good argument? and would it be? "Not even ONE PhD political scientist in America has refuted anything I have said in this thread." I'm embarrassed to make that argument even facetiously."

Actually, you should be embarrassed - but I doubt you are. I won't dignify it with any further comment.

Isn't there some pin head you need to be dancing on? ;)

--------------------------------------------------------------------

<Literally, I've just stumbled on most of these. I haven't even seriously investigated him yet. This was part of a post originally addressed to Febble>

Regarding Mitofsky:
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x371726#373177
(I've pasted the most important parts here:)
"...you revealed some interesting things about Mitofsky:

1) He claimed rBr as fact in his original report. At MOST, this was a hypothesis, he knew that, EVEN his rBr "supporters" call it that. Plain and simple, legitimate professionals NEVER claim a hypothesis as a fact - by definition, it is simply a "guess", or "hunch" that must be investigated further. THIS ACTION ALONE IS A STRONG BREACH OF ETHICS. Thanks for pointing it out.

2) As you also pointed out, he also advanced rBr WITHOUT ANY SUPPORTING EVIDENCE. Again, this is something no ethical professional would dream of doing, in a formal report, as he did. They would wait until the investigation was reasonably done, and ONLY THEN publish their findings. Again, Thanks for pointing it out.

In your post above, you say:
"I know almost nothing about him..."
When you find the time, it would be well worth your while to learn more about him. Here's some info to get you started:

---------------------------
Febble, would you at least listen to this interview? I've read your paper twice, printed it out, and highlighted it and notated it, so this isn't too much to ask, is it? ;)

OTOH, will YOU listen to this interview? I doubt it.

3. I've listened to Victoria Collier's interview, and highly recommend it

http://www.edwardsdavid.com/BushVideos/infidelguy.com_Vote_Scam_a.mp3
Also, read the first eight chapters (three more than at the votescam site) of the book here:
http://www.constitution.org/vote/

*<Verification here:>
http://www.votefraud.org/Archive/Write/newhampad.htm
"...early media reports stated that Bush and Buchanan were running neck and neck in the 1992 New Hampshire primary..."

<My original source:>
http://www.votefraud.org/News/2000/7/071800.html
-----------------------------------------------

And this is just the tip of the iceberg. Mitofsky has been at this since the mid-sixties.

I've found a clear pattern of <from another post of mine:>

http://www.democraticunderground.com/discuss/duboard.ph...
"Mitofsky uses a number of methods to "cover-up". It seems to range from computer meltdowns, "found votes", withholding the data, "convenient guesses", etc.

So what may have happened in 2004? <To "trip up" Mitofsky.>

- More scrutiny because of publicity of the 2000 theft.

- Internet leaks"

<Some other related links:>

Election Night Projections - Cover For Vote Rigging Since 1964?
http://www.ecotalk.org/VoterNewsService.htm

2002 Elections: Republican Voting Machines, Election Irregularities, and "Way-Off" Polling Results
http://www.ecotalk.org/MidtermElections2002.htm

Suspicion Surrounds Voter News Service
http://www.ecotalk.org/VNSclosingdoors.htm

Could the Associated Press (AP) Rig the Election?
http://www.ecotalk.org/AP.htm

Votes Downloaded to AP - gets direct feed from voting tabulating computers
http://www.ecotalk.org/APvotes.htm

<You know, it was YOUR assertions of how "squeaky clean" Mitofsky is (in Kathy Dopp's thread the Thursday before Mitofsky's presentation) that tipped me off to check him out. (I was getting really "bad vibes" by what you were doing in that thread, so I thought I better check him out.) I really didn't have much of an opinion on him, either way, before that (just a mental note to eventually check him out, because of his appearance in the Jan 19th Nightline show). Once I started checking him out that Thursday, the "dirt" just started pouring in!

I'm curious, do all three of you (OTOH and Bruce) generally hold Mitofsky in such high regard? If so, is there a pattern here?>

[OnTheOtherHand]

293. stand back! he's using the power of animated emoticons!

"Uncle Warren"?

Wow. You are nothing if not prolific.

Tom, what if anything are you actually trying to accomplish here? How are you going to blow up Mitofsky without blowing up the exit poll evidence? Heck, if you can prove that the guy has been rigging elections for decades, let us know.

[tommcintyre]

306. I accept your capitulation, and here's a chance to redeem yourself...

But first, on the use of emoticons: In my studies on interpersonal communication, I recall estimates that as high as 93% is nonverbal. Isn't our "bare" writings (as we are doing here) very similar to the verbal element in a face-to-face conversation? So, any "helpers" we can use, such as: "quotes", bolding, ALL CAPS; and, yes, even emoticons - animated, or not - of value in getting our ideas across?

So, OTOH, try "thinking outside of the box" you appear to be in.
-------------
Now on to your apparent capitulation; and opportunity to redeem yourself:

First, I assume by your "silence" (avoidance of the issues in our current series of posts); you have acquiesced to ALL of them? (ODell says below that silence can be construed as capitulation - I tend to agree with him on this.)

So, GREAT! We are never too old to learn, are we? ;)

Now, on to your redemption opportunity:

Why don't YOU "blow up Mitofsky"? After-all, I believe you have professed a desire to ACTUALLY HELP in exposing the election fraud, haven't you? (And, it doesn't appear you understand the math too well, re: Kathy's post below.)

Well, here's your golden opportunity! I have provided you with the links/information to get you started; and, as I recall, you're even on a "first-name-basis" with "Uncle Warren" (you email him a few times a year).

First, you could pose a series of questions to him that, no doubt, many of would like answered. I can help you with the list of questions - even better - you can start a thread with a heading like "QUESTIONS FOR WARREN MITOFSKY". This way, others can participate too - very democratic, don't you think?

Now, just in case I have misinterpreted you, and you are saying something like: "Hold on there pardner! I believe Uncle Warren is as clean as the driven snow!"

Hey... no problem! His answers to these questions could be used to exonerate him! <Time for the sarcasm emoticon again? ;) > :sarcasm:

Well... one way or another, I think this thread needs to be made - so, if you don't do it, I will.

Once we have compiled the questions, will YOU email them to Mitofsky for us? That would be very nice (and helpful too).

Also, do you think we could post those questions for him on Mystery Pollster (I understand Warren posts there sometimes)? That may be useful too (to get to the bottom of all this).

Please let me know your thoughts on all this.

---------------------------------

Oh, finally...

You said:
"How are you going to blow up Mitofsky without blowing up the exit poll evidence?"

Well, I guess you missed it in the post you just replied to:

"So what may have happened in 2004? <To "trip up" Mitofsky.>

- More scrutiny because of publicity of the 2000 theft.

- Internet leaks"

Here, I'll explain it for you a little more:

Normally, Mitofsky "fixes" ("adjusts") the American exit polls to match the election results. THIS IS CONSIDERED VERY QUESTIONABLE (see link below).

http://en.wikipedia.org/wiki/2004_U.S._presidential_election_controversy%2C_exit_polls
"Because final published exit polls in America are matched to vote counts, they cannot be used to determine election fraud. However, in the 2004 election, pre-matched exit polls were leaked onto the internet."

Some related information:
http://en.wikipedia.org/wiki/Exit_poll
"Exit polls have historically and throughout the world been used as "parallel vote tabulation", as a check against and rough indicator of the degree of fraud in an election."
snip
"Leaks of exit poll figures for the 2004 presidential election, mainly via the Internet, appeared to indicate a victory for John Kerry. The discrepancies between the exit poll data and the vote count that where outside of the margin of error, coupled with irregularities in the election which seem to explain the discrepancies and what many perceive as evasive tactics by the polling companies, have shed doubt on the legitimacy of that election among political activists and government officials."
<Are you ready to "blow him up" yet, OTOH? Am I being "prolific" again. ;)>
------------------
So, how did we catch him this time?

Again:

- More scrutiny because of publicity of the 2000 theft.

- Internet leaks

OTOH, have you listened to Victoria Collier's interview yet? You really need to.

3. I've listened to Victoria Collier's interview, and highly recommend it:

http://www.edwardsdavid.com/BushVideos/infidelguy.com_Vote_Scam_a.mp3
Also, you can read the first eight chapters (three more than at the votescam site) of the book here:
http://www.constitution.org/vote/

I'll be asking you again about this.

[OnTheOtherHand]

340. what do you think I capitulated in?

I'm an assistant professor (which means I'm tenure-track)? sure, I capitulate in that.

My field is American politics? sure, I capitulate in that.

If no Ph.D. mathematician critiques the paper, you will take that as proof that the paper is good -- and if one does, you will take that as proof that he or she is a hack? sure, I capitulate in that.

You sure know how to use bold-face? sure, I capitulate in that. Hey, go wild, guy.

[tommcintyre]

342. Uh, OK. I guess I "misunderestimated" you.

Unlike ODell, your silence on the issues equals continued evasion and obfuscation?

And, uh... your one er... <sort of> unambiguous statement:
"If no Ph.D. mathematician critiques the paper, you will take that as proof that the paper is good -- and if one does, you will take that as proof that he or she is a hack? sure, I capitulate in that."

And, WHO exactly IS this "Ph.D. mathematician" who has "critiqued" the paper? I missed that.

Ah... you don't mean YOU now do you? If so, WHERE, except in your claims here, are you LISTED as a mathematician? You majored in "American politics?" (now THAT'S a field you can "take to the bank" outside of the academic arena ;) ).
So, did you get a degree where you minored in mathematics (at least)? If so, wouldn't you AT LEAST list THAT under your list of "interests"?

Oh, and one other thing: You didn't mention anymore about Mitofsky in THIS post. So does this mean you are considering taking on the task of "blowing him up" as you put it. Or, maybe even exonerating him, as I quipped. Or, have you just gone into "evasion mode" on this issue too?

I'm just wondering...

[OnTheOtherHand]

347. if anyone else has questions about this...

feel free to PM me, and I will explain. Folks can also poke around and judge whether I have been silent on the issues.

Tom, you can PM me too, if you are actually interested in my answers.

[sunshinekathy]

247. Kathy Dopp again - slight correction

Sorry, but it was late last night when I posted and a trucker ran into my parked trailer and I was tired and exagerated slightly.

Bruce's math is not "100% wrong". His math may be correct as math, but not set up in a way that would truly describe the exit poll discrepancy data at hand. I have not had time to read his work in detail, but all his emails to me showed that he did not fully understand the work of Ron Baiman or myself and was mischaracterizing what our working paper said.

Bruce's conclusions are wrong. Mitofsky himself does not even make the claims that Bruce does.

Many many good things are happening for US Count Votes and we will get beyond this.

American patriots can succeed in ending the embezzlement of votes that has been caused by decades of not auditing electronically and mechanically counted votes, and we will.

Best,

Kathy

[RonB]

251. High Bush Precincts Show Statistically Anomalous K Values + More on ODell

Dear DU Readers,

The following is an analysis that compares the O'Dell simulation data to the USCV ”aggregate analysis” (See Appendix A, Table 3, p. 25, of second April 12, USCV Report) and investigates the general question of statistically significant differences in K and B (Table 1 below). For those who are interested I will be glad to provide the spreadsheets with the complete analysis.

The summary conclusions are:

1) Weighted average K is 57.8% and weighted average B is 50.4% for the O’Dell data. This is consistent with the aggregate analysis showing that rounded K must be at least 58% to B's 50% to explain the E-M data. The E-M K=56% and B=50% “hypothetical” cannot explain the E-M data.

2) There is very little difference between the O’Dell data and the USCV Aggregate data. In particular the "unconfounded" variables of interest (the partisan response rates K and B) are almost all within 1% of each other except for the high Kerry (K mean difference of 2.1%) and high Bush precincts (K mean difference of 7.1%). The medians are within one percent in the former but deviate by 5.5% in the later.

3) These small differences may reflect the fact that the O’Dell simulation does not exactly hit the E-M mean and median, values in every case, whereas the aggregate results do so by design.

4) The discrepancy may also be due to unexplained changes in variance across quintiles for the O’Dell simulation data. These are 2.25, 2.87, 4.60, 4.50, and 2.47 times "maximal random variance" for K (and roughly similar for B), across quintiles respectively. If this reflects E-M's variances it needs to be explained! At a minimum, the simulation variances do not seem to adjust appropriately to sample sizes.

5) The discrepancy may also reflect the unexplained pattern of increasing linear correlation between R and E across quintiles in the O'Dell simulation data: -0.64, -.30, 0.15, 0.47, 0.90. Since K = (k-0.5E)R/k, a positive R and E correlation will reduce K (and alpha since B is fairly constant - see below) and a negative R and E correlation will increase alpha. This may explain why the simulation mean alpha is significantly higher than the aggregate mean alpha in high Kerry districts and significantly lower in the high Bush districts. This unexplained correlation pattern thus serves to "flatten" alpha.

6) Finally, the crux of the matter.

For each precinct, a perfect unbiased random sample would give K=B=R where R is the overall exit poll participation rate in that precinct. However, although, average R across quintiles ranges from 52% to 56%, R undoubtedly differs considerably from precinct to precinct. Since the exact R for a given precinct is unknowable, I conservatively set R=0.5 across all the precincts so that the standard deviation of K and B will be at its maximal possible level of “square root of (0.5x0.5/n)” for a sample of n precincts (these are the SD of K values calculated in Table 2 below).

Using these maximal standard deviations for K and B for each quintile, I perform standard tests of the differences of the means of two proportions (see any Statistics text for the formulas for “random stddev of K2-K1”, and normal probability table – or exel spreadsheet “normdist” function- for “prob of K2-K1 if K2=K1”).

The results (whether calculated from aggregate K and B derived from means, O’Dell simulated mean K and B, or aggregate K and B derived from medians – to remove outlier effects) all show that high Bush precincts must have significantly higher Kerry exit poll response levels, than almost all other districts.

The probability that the high Kerry exit poll response levels necessary to generate the very large mean and median WPE errors for these high Bush (87% Bush reported vote) precincts, could have occurred “randomly”, if these precincts had “true” K values equal to those of the other precinct categories, is very small. These probabilities are almost all less than 1% (one tail) for K calculated from aggregate means (see Table 1, Appendix F, USCV report). For the O’Dell simulated means they are from 4% to 9% (in spite of the problems with the O’Dell simulation data noted above). Even for K calculated from aggregate medians (see Table 1, Appendix F) they are mostly (except for “moderately high” Bush precincts) from 5% to 10%. For O”Dell simulated medians these probabilities are from 1% to 4%. Note that these probabilities have been calculated from “maximal” random errors. More accurate random error calculations (for which precinct level data is required) would result in even smaller levels of significance.

In stark contrast, B levels do not appear to be significantly across precinct categories, no matter what kind of data is used. This indicates that “bias” (K/B) goes up significantly in high Bush precincts. The difference between calculated K and B levels across precinct categories is striking.

These findings show that (high Bush) precinct partisanship is strongly linked to a (calculated) unexplained increase in Kerry voter exit poll response. A linear correlation between “alpha” (Ln(K/B)) and precinct partisanship across the entire data set is irrelevant to this finding (see previous posts).

Moreover, the large variance in individual K and B levels for individual precincts across all partisanship categories is also statistically irrelevant. We know that sample data (almost always) includes large levels of “measurement” error (beyond random sampling error). What is important is whether the averages or medians of these measurements are biased in one way or another. These means and medians are the statistically meaningful parameters, not the individual measurements. We find that they are indeed biased, in one direction, and for a particular category of partisan precincts.

This is an important finding. Something went on in these high Bush precincts that "jumps out" statistically. This finding is consistent with the Bsvcc (“Bush Strongholds have more Vote-Count Corruption”) hypothesis discussed in all three USCV reports. It is inconsistent with the constant mean “bias” (K/B) "hypothetical" offered by E-M (now refuted with regard to having a "constant mean" by Mitofsky), though it was seemingly supported by the zero linear correlation finding presented at AAPOR.

If this unexplained calculated increase in Kerry voter exit poll response did not actually occur so that exit poll results for these precincts showing higher Kerry margins were correct, the high Bush reported election results for these precincts were mis-tabulated.

A serious (statistically rigorous multi-factor regression) explanation explaining these discrepancies has not been provided. The “constant mean bias hypothetical” has been shown to be inconsistent with the data (and refuted by Mitofsky). The data that would allow independent analysts to see if an exit-poll explanation can be found for these anomalies has not been provided.

Given the massive on-the-ground evidence of voting irregularities (circumstantial and other – including recent evidence of code in Diebold memory cards specifically allowing for vote corruption – see http://www.scoop.co.nz/stories/HL0307/S00064.htm ), it is reasonable to conclude that vote corruption is a viable alternative hypothesis for explaining these discrepancies.

7) Conclusions: a) there is hardly any difference between the O'Dell simulation data and the "aggregate" analysis except for unexplained correlations and variances that are included in the O’Dell simulation analysis, b) there are highly significant differences between the K values in the high Bush precincts and all other precincts - just as USCV has hypothesized in all three of its reports - again no matter how the analysis is done.

Best,

Ron

P.S. - I've reproduced one of the tables below - the others are similar with different K and B values entered. All show statistical significance for K differences in the bottom row.

Table 1: Two Sample Differences of Proportion Averages for K
Calculated from Aggregate Means

k=.87 k=.67 k=.5 k=.33 k=.13

1 2 3 4 5
N= 90 165 540 415 40
K= 0.529 0.574 0.564 0.601 0.775
SD of K= 0.053 0.039 0.002 0.025 0.079
R= 0.53 0.55 0.52 0.55 0.56

1 90
0.529
0.053
0.53
2-1 =quintiles compared
2 165 0.045 =K2 - K1
0.574 0.558 =K0 - if K2=K1
0.039 0.065 =random stddev of K2-K1
0.55 0.2446 =prob of K2-K1 if K2=K1

3-1 2-3
3 540 0.035 0.01
0.564 0.559 0.566
0.002 0.057 0.044
0.52 0.2679 0.4103

4-1 4-2 4-3
4 415 0.072 0.027 0.037
0.601 0.588 0.593 0.580
0.025 0.057 0.045 0.032
0.55 0.1042 0.2752 0.1254

5-1 5-2 5-3 5-4
5 40 0.246 0.201 0.211 0.174
0.775 0.605 0.613 0.579 0.616
0.079 0.093 0.086 0.081 0.081
0.56 0.0041 0.0096 0.0046 0.0153

[OnTheOtherHand]

252. a few points

I haven't yet grasped every aspect of Ron's statement, but I see several things I can agree with and some that I at least partly disagree with.

Among other things, I agree that the true "alpha" must be closer to 1.15 (in the numbers above, 57.8%/50.4% is about 1.146) than 1.12 (56%/50%) in order to match the E/M results.

When I entered alpha = 1.15 and a standard deviation for response rates of 12% into my own simulator, my third run yielded mean WPE = -9.99, median = -5.84, and mean abs WPE = 13.2. That is actually too close to the observed values; of my 100 runs, only 4 yielded mean WPEs of -9.5 or "more." Still, I continue to fail to grasp Ron's confidence that the numbers are beyond the bounds of plausibility. (Part of the difference may be that I model sampling error at the level of the individual respondent.)

Looking back at O'Dell's numbers and at the E/M spreadsheet, a handful of outliers (three or four) account for most of the statistical quirkiness in the high-Bush precincts. (As I've just noted, these outliers can be generated randomly, but in real life they could have various benign or malign causes.) This fact begs the questions: do a handful of outliers in high-Bush precincts support an inference of massive fraud? what inferences do they support? what other findings would or would not support an inference of massive fraud?

O'Dell's paper is a good place to look for further insight on these questions -- surely not the last word.

[RonB]

253. Problems with the Four Outliers "Explanation"

Dear "OnTheOtherHand":

The following is a copy of an email I just sent responding to the issue of the "four outliers".

a) 4 out of 40 is 10%. 10% of high Bush precincts nationwide is a large number of precincts! (Also, I'm still not clear why there are only 40 high Bush precincts in the sample versus 90 high Kerry - I suppose it could just reflect over sampling of high Kerry precincts, but the very large difference in samples sizes needs to be explained - were some other outliers precincts dropped?)

b) Calculations from medians are also significant for partisan precinct quintile comparisons between precincts 5-1, 5-2, and 5-3, from the 5% to 10% one-tailed level for these "maximal" standard deviations. This suggests that its not just a matter of a few outliers.

c) Even if this is a result of a few outliers - though b) suggests its not - you can't just toss out the outliers unless they are really anomalous - but the scatter plot suggests that there are large outliers in all categories of precincts. This suggests pervasive across-the-board measurement error (or vote corruption). This is almost always true of survey data (that is large outliers and measurement error). The key is (almost always unless a clear reason can be found for the outliers) to look at means and medians and their statistical significance and not individual data points - is the average of the outliers biased in one direction or the other? Here they clearly are and its statistically significant. This should not be dismissed. It points to a need for an on-the-ground investigation of (especially) high Bush precincts.


E-M seems to believe that they have an exit poll explanation for the discrepancies. They need to understand that they could be wrong. They need to allow independent analysis to look at the data and at whatever serious (multi-factor regression) analysis they have done.
To date neither of these has been released.

The more we (independent analysts) look at the data that has been provided the more strikingly urgent this appears!

Best,

Ron

[OnTheOtherHand]

254. sure, let's consider these points

Folks generally call me OTOH over here, to save typing. And if I get in trouble, they can yell, "Run, OTOH, run!"

a) In the E/M tabular analyses, 40 of the 1250 precincts are high-Bush, about 3%, but some differential deletion is quite likely, so the true figure might be closer to 6%. (I doubt it is higher than that, because I would expect more high-Kerry than high-Bush precincts -- but extreme precincts on both ends might be disproportionately truncated.) 10% of 6% would be 0.6%. We're talking about WPEs on the order of 60% in favor of Bush, so a back-of-the-envelope estimate would be a 0.36% swing -- assuming that these precincts reflect fraud rather than exit poll mess-up. That's both a pretty freehand extrapolation from a handful of precincts, and a long way from accounting for the exit poll discrepancy. The original question we set out to answer, I thought, was "Why such a large difference between the exit polls and the official returns?" The outliers in high-Bush precincts could hardly begin to account for the difference even under extreme assumptions.

I would expect more high-Kerry than high-Bush precincts because I have rarely encountered any Republican analogue to the solid-Dem precincts I encountered in central Ohio. (I'd like to see a nationwide precinct partisanship distribution based on official data from some presidential election, perhaps 2000.)

The observed results could nonetheless be consistent with widespread vote embezzlement across the range of partisanship -- a signal. But so far I have seen no good evidence that they are. For reasons explored by Liddle and discussed by O'Dell, I would expect widespread vote embezzlement to induce a statistically significant correlation between partisanship and ln alpha; we observe none.

b) This result seems to hinge on your finding that the median WPE of -5.8 in high-Bush precincts entails a higher K response rate than either the mean WPE of -10.0 in the same precincts, or the higher median rates in two other partisanship ranges. The latter finding is plausible (although probably sensitive to specific assumptions about how partisan the high-Bush precincts are), but the former requires explanation.

Your "maximal standard deviations" are no such thing, as evinced by the fact that the O'Dell simulation yields much larger standard deviations (if I understand you rightly, which I may not) and yet falls short of the reported mean absolute WPEs. I don't know what they are, but I don't consider them a viable basis for statistical inference.

In the O'Dell simulation, the median ln alpha for the middle precincts is 0.169. The median ln alpha for the high-Bush precincts is either 0.193 at Bush=83%, or 0.226 at Bush=87%. (You can check those figures -- I don't know whether different versions of the sim are in circulation.) While that latter figure begins to look dramatic, the bootstrapped standard error of the median is around 0.08 (the standard error of the mean is over 0.10), so the difference in medians does not even approach statistical significance.

c) I don't see what statistical analysis you have performed comparing high-Bush to high-Kerry outliers or considering what assumptions are necessary to generate those outliers. If your only basis for inferring statistical significance is the maximal standard deviations, then you are basically nowhere.

I certainly hope that E-M will carefully investigate its extreme outliers (if it hasn't already) and take whatever action it thinks is appropriate. Nevertheless, I think it is important to say clearly that I see no statistical evidence supporting fraud. The results we have discussed seem consistent with the null hypothesis: a sloppy, noisy survey with pro-Kerry response bias. (I also see no statistical evidence strongly infirming widespread fraud. I don't think one can infer much of anything from these data.)

I agree that E-M need to understand that they could be wrong, that they should release serious analytical work, and that some provision should be made for direct analysis. If blurring is viable, it will take money.

In my case, the more I look at the exit poll data, the less strikingly urgent that particular part of the election reform struggle appears. But the trend may yet reverse.

[Melissa G]

256. Hi OTOH, If you have the data, pm Dzika... he might oblige...

you wrote..
I would expect more high-Kerry than high-Bush precincts because I have rarely encountered any Republican analogue to the solid-Dem precincts I encountered in central Ohio. (I'd like to see a nationwide precinct partisanship distribution based on official data from some presidential election, perhaps 2000.)

We all might enjoy looking at this map. (and BTW those high K precincts we would be looking at were part of the reason of the Ohio win for K....)
:hi:

[OnTheOtherHand]

267. ah, if I had the data...

I would do it myself, at least the descriptives -- but I would like to see the map too. (Actually, a map of every precinct in the U.S. would be pretty hard to work with, but the ability to "drill down" from the county level into wards/districts and precincts would be excellent.)

The National Election Data Archive is a really great idea. Some folks did something called "Record of American Democracy" that pulled together election results for umm I don't remember (you could google it if curious), but maybe a decade or so. It's tough work. Although it actually should be getting easier because more of the raw inputs are already digital.

[LightningFlash]

259. Your issuance of generalities does not explain the underlying factor.

Bush would have needed 55.67% of united states refusers, in order to even complete the data or come near the exit poll discrepancy. He would have needed it in only eastern states.

With the final facts being the last, adjusted weighting has proven bogus how can you possibly assume that there is virtually any validity to the hypothesis?

The gender weightings reversed, with the female vote becoming smaller

1.5 million of those bush voters who voted in 2000, were included, even though they had died(at minimum)

The switching of weighting on the religious chard ID can not be explained (as per http://exitpollz.net )

You are no closer to getting underneath any of these bogus demographics, than you are to saying the exit polls were useless.

The exit polls in Florida state have, just the opposite, proved fraud with the analysis completely backing it up.

""Miami-Dade County, Florida. May, 2005. ES&S iVotronic.
New evidence shows both phantom votes and lost votes in the November election.
The number of voters reported by election workers didn't match the number of ballots cast in 260 (35%) of Miami-Dade's 749 polling places. Some showed more votes than voters ("phantom votes"); others showed significantly more voters than ballots cast.*

While some of the discrepancies can be traced to sloppy procedures and training, others are evidence of problems not yet explained: "


Phantom votes: precinct with a major difference was Precinct 362, which recorded 583 votes and 859 signatures. A review of that precinct's signature log found 580 signatures.
Phantom votes: in Precinct 41 there were 910 votes and 844 signatures, a difference of 62.

Lost votes: At combined Precinct 117/166, the ballots totaled 995 and the signatures numbered 1,276.

Phantom votes ... then more phantom votes: For Precinct 816, in the Church of the Ascension at 11201 SW 160th St., a Review inspection of the voter log showed 945 signatures, while the iVotronic computer tape showed a count of 1,032. But the individual machine counts are listed on the tape as well, and they add up to 945. But the number certified by the canvassing board came in at 1,116 votes.

In this case, the discrepancy may have been due to an iVotronic machine malfunction. Lynn Kaplan, a volunteer observer for the reform coalition who was at that polling place on Nov. 2, said in an interview that as a poll worker was closing down one of the iVotronic machines at the end of the day, an error message popped up on the machine's digital screen saying: "Internal malfunction/unit closed to save data/vote data corrupted."

... said the situation in Precinct 816 highlights the need for better procedures and training for reconciling signature and ballot totals. She also stressed that the county must investigate discrepancies to see if they resulted from fraud or equipment malfunction.


The Secretary of State's office has refused to get involved.

Jenny Nash, a spokeswoman for Florida Secretary of State Glenda Hood, said the issue of reconciling signature totals and machine counts was a local one and is not the concern of her office. "Each supervisor has their own process for how they reconcile the numbers," Nash said.

* Elections Discrepancies found in 35 percent of Miami-Dade precincts. Daily Business Review. May 06, 2005 By: Jessica M. Walker;"

http://www.dailybusinessreview.com/news.html?news_id=34733 "
http://www.votersunite.org/info/content/newmessup-17.asp

http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01008-machine7301.PDF
http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01006-machine6545.PDF
http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01004-machine21285-oct.PDF
http://www.bbvdocs.org/FL/PalmBeach/Nov2004/pollsite01004-machine21284.PDF

Proven wrong. Over and over again. And not one single point is responded to!

"How can the discrepancy actually be that large?"

Not unless the exit polls had a miracle, and trained everyone to do computer-laser accurate oversampling of voters everywhere and did it so well it moved mountains.

Or, the valid and final explanation: Machine manipulation. Which can and DID vary the discrepancy completely in the state of Florida, where there was only about 10% uniform random bias...

[OnTheOtherHand]

268. fraud vs. exit polls

Did you notice that not a single thing you copied and pasted about Florida substantiated or even addressed your claim that the "exit polls in Florida state have, just the opposite, proved fraud"? If you think that the things you copied and pasted about Florida prove that Bush stole the state, we can discuss that.

If the exit poll was wrong, then I would expect some of the weighted demographics to be wrong as well. There is no magic formula for extracting good results from bad data. I don't know whether the ones you have cited are demonstrably wrong. (I actually did some work to explain why the preponderance of recalled Bush voters vs. Gore voters didn't seem so strange. I haven't done this work for other criteria; I don't think I'm obligated to be more curious about every single argument than the people who make the arguments.) But even if they are, it would not settle whether the 'raw' exit poll was right or wrong.

[LightningFlash]

270. Your response.

"we can discuss that"

Really can we? Then why have you not addressed one single item? Bev Harris's own "infamous" Black Box Voting did all the audit reports, coming from three or more of Florida's largest counties filled with the "phantom" voting.

:thumbsdown:

[OnTheOtherHand]

273. umm, dude...

You might refer back to the top of the thread and see what this thread is about.

If Bev Harris has proven fraud in Florida, then Bev Harris has proven fraud. If Bev Harris has proven that fraud is even possible, then we all should be howling about that. Instead, folks are calling for Bruce O'Dell's head on a stick because he had the temerity to disagree with Dopp's interpretation of the national exit polls. And some of them say that I am wasting time. Bizarre.

If you agree that O'Dell has no right to criticize their interpretation of the national exit polls in public, then defend that premise, please. Otherwise, it might be appropriate to agree that O'Dell has that right, that his critics should back off, and that we should get on with business. When you insist on changing the subject, I am not the one ignoring points.

[LightningFlash]

277. This is frivilous..

And it proves exactly what I had been saying all along, about the "Right Wing" talking points and not living in standard reality...

Regardless of what anyone has thought about Bev Harris, her organization has conclusively demonstrated fraud and hacked machines in the biggest counties, including a tampered with Optiscan.

Reading side by side with the exit poll chart, it confirms the analysis that the exit poll was largely correct and the discrepancy can be attributed to vote manipulation.

Vote manipulation by "ES&S" and democrats in name only? Most absolutely, if you consider the one opposite of Constance Kaplan, who is called Theresa LePore, is a certified DINO. In fact, it was proven the memory cards had secret code planted on them. If this wasn't anymore black and white there could be pen-writing on the wall.

The "point" we come back to here is, the standard deviation bias was not more than 8% in Florida. What "evidence" is there to even suggest it was more than that, or strongly "reluctant" in all the other states?

Conclusively reluctant responders can be shown on both sides of the aisle, but Kerry voters even, have been shown to be more reluctant in such demographic "hypothesis" and no one even considers that point.

Nowhere in O'Dell's analysis is he considering the fact there is at least equal chance, if not more, for Kerry responders especially republican to be reluctant. And they are missing or omitted from the final data, which is a FURTHER RED FLAG STILL STEPPED OVER BY BRUCE O'DELL.

And he claims this whole disagreement was over methodology. If that was the case, lest not forget the official "census" 2004 data has proven to be the most reliable methodology.

http://www.census.gov/population/www/socdemo/voting/cps2004.html

You can not disprove any exit polls, unless the basis for such fact is well stemmed in reality. So far it isn't, even when taking into account the fact there was both bias and fraud. :nuke:

[Name removed]

278. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

280. Deleted message

Message removed by moderator. Click here to review the message board rules.

[OnTheOtherHand]

282. is there an argument here?

It won't do to cite MoEs for the overall turnout rate, if the Census Bureau says they are inflated.

Are you talking about a difference in turnout rates between some group and some other group (which might well hold up despite the overall inflation), and if so, which groups?

Are you talking about the apparent proportion of women in the electorate in the CPS, as compared with the apparent proportion in the weighted exit polls?

I haven't accused the Census Bureau of making up bogus weightings. I haven't even accused E/M of making up bogus weightings, although the weightings could easily be wrong. As I said, there is no magic formula for extracting good conclusions from bad data.

I don't think any of this has to do with O'Dell, but it would at least be salient to the 2004 election, which is good.

[LightningFlash]

291. Your logic runs in circles.

The U.S. government conducted the census, if you disagree with them take the issue up over there. Otherwise, there doesn't hold any credability to what you're saying.

And for the final time, the weightings must be bogus or "faulty" if the exit polls are faulty. So the weightings are faulty no matter what, which you apparently can not explain the adjustments in gender, voted in 2000, or other without vying that explanation.

Although several state exit polls have backed up the data at hand on confirmed vote manipulation you fail to address anything other than call the "Census" statistics uncredible, of which there is close to 3.5 million votes not accounted for. End of comment.

[OnTheOtherHand]

292. second verse, same as the first...

(1) It wasn't a census. It wouldn't be any harder to get this right, so why not?

(2) The Census Bureau states that the CPS overestimates turnout. I agree with them. Do you?

(3) I am happy to agree that the weightings are faulty no matter what -- I just don't see where that gets us. The rest of the sentence I don't understand. What do you think the CPS data prove, and why? Don't just post the link again, make an argument.

(4) Let's see the quantity of vote manipulation we can demonstrate, and then -- blessed joy -- we might never have to mention the exit polls again.

[LightningFlash]

295. Already showed the quantity based on a state. Guess you ignored this.

[OnTheOtherHand]

298. what state can you prove was stolen? this is important. n/t

[LightningFlash]

301. Florida. Keep repeating and stressing the same thing.

Florida, Florida, Florida.

http://www.tampabays10.com/news/news.aspx?storyid=14806

Exit poll (STATE EXIT POLLS) plus leaked results out of the three largest counties, make up the discrepancy universally despite bias. Broward, Palm Beach, Volusia, Miami-Dade Counties.

It only took that, and swung it the other way. Democrats in name only installed Diebold there.

[Internut]

302. "Proof" is not the same as "allegation "

those are two different words and two different concepts. I know on this forum the two are equated all the time, but that does not make it correct.

[autorank]

303. Seriously, are you serious?

There are various meanings of "prove." The one that counts ultimately is a legal standard of "proof." That means you have enough information for an investigation, to start, and that the investigation produces a prime efacia case, one that can stand on its own. We've got that without much doubt, do you agree? Once the case is argued, the judge and/or jury decide which side proved its case and the decision is rendered as a result of one or the other side proving its case.

To get to legal "proof" by "proving" that this or that state or the national election was stolen is another conundrum. It cannot be done.

There is more than enough here by TIA and elsewhere to warrant a robust investigation and enough here and elsewhere, in my opinion, to establish a prime efacia case and move to trial.

Are you arguing at the gates of this path that there isn't even enough evidence for a robust investigation? Now, this IS important.

Contact the DNC and Give 'em Hell About NOT Acting on Election Fraud

NEW LEADERS FOR A NEW DEMOCRATIC PARTY

[OnTheOtherHand]

348. no, I wasn't opposing investigation -- not my point at all. n/t

[TruthIsAll]

262. The Exit Poll Response Optimizer should tell you something.

Here is a run of the Exit Poll Response Optimizer for what I
believe to be the most likely and plausible scenario.

INPUT
1. Target: Kerry wins 52.5% of 2-party vote
2. Response = R = 53% (wtd average)
3. Alpha = K/B =.98 (wtd average)
(Kerry voters slightly more reluctant 49 Kerry/50 Bush)
I believe Repubs for Kerry were intimidated. 
It makes more sense intuitively AND anecdotally.
E-M says that Alpha = 56/50 =1.12 


OUTPUT:
4. Kerry wins Exit Poll with 51.46% of 2-party vote

5. Total 2 party vote (121mm): 
Kerry 63.52mm, Bush 57.47mm 

6. Avg WPE: 2.08%

7. Exit Poll Response:
Ranges from 48% (High Bush precincts) to 52.5% (Non-partisan)
to 58% (High Kerry).


EXIT POLL RESPONSE OPTIMIZATION MODEL 							
6/9/05 10:21 AM							
							
Objective: Determine values of constrained variables required
to derive a target Kerry/Bush percentage split using aggregate
exit poll response data.							
							
Precinct Variables Range Constraints:							
1-Response: within (Min, Max) and equal to weighted avg
rate.							
2-Kerry win percentage:within (Min, Max).							
3-Alpha (K/B): within (Min, Max) and equal to weighted
average.							
4-WPE within E-M actuals or (Min, Max). 							
							
TARGET INPUT	MOST LIKELY SCENARIO						
Kerry 2-party vote		52.50%					
Bush 2-party vote		47.50%					
							
Wtd Avg Response		53.0%					
Wtd Avg Alpha (K/B)		0.98					
							
RESPONSE INPUT CONSTRAINTS							
							
1250	Strong Bush		Strong Kerry			
Prcts	40	415	540	165	90		
							
KERRY WIN%							
Min	0%	20%	40%	60%	80%		
Max	20%	40%	60%	80%	100%		
							
RESPONSE							
Min	30%	30%	30%	30%	30%		
Max	70%	70%	70%	70%	70%		
							
ALPHA							
Min	0.10	0.10	0.10	0.10	0.10		
Max	10.00	10.00	10.00	10.00	10.00		
							
WPE							
E-M	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		
Min	-10.0%	-10.0%	-10.0%	-10.0%	-10.0%		
Max	10.0%	10.0%	10.0%	10.0%	10.0%		
							
OPTIMIZER OUTPUT SUMMARY							
							
	PERCENTAGE SPLIT	VOTE SPLIT (mm)				
PCT	POLL	VOTE	Diff	POLL	VOTE	Diff	
Kerry	51.46%	52.50%	1.04%	62.27	63.52	1.26	
Bush	48.54%	47.50%	-1.04%	58.73	57.47	-1.26	
							
Bush needed	46.35%	of refusers to match his vote					
							
PRECINCTS						Wtd Avg
Number	40	415	540	165	90		
Weights	3.2%	33.2%	43.2%	13.2%	7.2%		
							
Resp.	48.0%	51.1%	52.5%	58.0%	58.0%		53.0%
Dev	-5.0%	-1.9%	-0.5%	5.0%	5.0%		0%
							
ALPHA							
K/B	1.19	1.00	0.93	1.04	0.96		0.98
Dev	21.8%	1.8%	-4.6%	6.3%	-1.9%		0%
							
2-PARTY VOTE 							
Kerry	19.2%	40.0%	54.8%	60.2%	96.7%		52.50%
Bush	80.8%	60.0%	45.2%	39.8%	3.3%		47.50%
Votes 							
(mm)	3.87	40.17	52.27	15.97	8.71		121.00
Kerry 	0.75	16.08	28.66	9.62	8.42		63.52
Bush	3.13	24.10	23.61	6.35	0.29		57.47
Diff	-2.38	-8.02	5.05	3.27	8.13		6.05
							
EXIT POLL							
Kerry	21.9%	40.0%	52.2%	61.8%	94.3%		51.46%
Bush	78.1%	60.0%	47.8%	38.2%	5.7%		48.54%
Votes 							
(mm)	3.87	40.17	52.27	15.97	8.71		121.00
Kerry 	0.85	16.05	27.28	9.87	8.21		62.27
Bush	3.02	24.12	24.99	6.10	0.50		58.73
Diff	-2.17	-8.07	2.30	3.77	7.71		3.53
							
WPE	-5.4%	0.1%	5.3%	-3.1%	4.8%		2.08%
							
							
							
OPTIMIZATION MODEL							

Categ.	HighB	Bush	Even	Kerry	HighK	.	Total/Avg
Prcts	40	415	540	165	90		1250
Kerry	20%	40%	54%	61%	96%		52.14%
							
ALPHA							
K/B	1.19	1.00	0.93	1.04	0.96		0.98
AvgDev	22%	2%	-5%	6%	-2%		0%
							
RESPONDERS							
Total	19	212	283	96	52		663
Pct	48.0%	51.1%	52.5%	58.0%	58.0%		53.00%
							
Kerry	4	85	148	59	49		345
Pct	21.9%	40.0%	52.2%	61.8%	94.3%		51.46%
Bush	15	127	135	37	3		317
Pct	78.1%	60.0%	47.8%	38.2%	5.7%		48.54%
							
REFUSERS							
Total	21	203	257	69	38		587
Pct	52.0%	48.9%	47.5%	42.0%	42.0%		47.00%
							
Kerry	3	81	148	40	38		311
Pct	16.8%	40.1%	57.7%	58.1%	100.0%		53.65%
Bush	17	122	108	29	0		276
Pct	83.2%	59.9%	42.3%	41.9%	0.0%		46.35%
							
VOTE							
Kerry	8	166	296	99	87		656
Pct	19.2%	40.0%	54.8%	60.2%	96.7%		52.50%
Bush	32	249	244	66	3		594
Pct	80.8%	60.0%	45.2%	39.8%	3.3%		47.50%
							
WPE							
Kv-Bv	-61.5%	-20.0%	9.7%	20.5%	93.3%		5.00%
Kp-Bp	-56.1%	-20.1%	4.4%	23.6%	88.5%		2.92%
							
WPE	-5.4%	0.1%	5.3%	-3.1%	4.8%		2.08%
E-M WPE	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		-6.77%

[TruthIsAll]

263. And so should this alpha sensitivity analysis

11 ALPHA SCENARIOS: BASE CASE (NO BIAS) ALPHA = 1.0 
 
Edited on Thu Jun-09-05 08:13 AM by TruthIsAll
 
			ANALYSIS OF EXIT POLL RESPONSE				
			BY PARTISANSHIP FOR VARIOUS ALPHA (K/B)
	             AND CONSTANT RESPONSE RATE (R=53%)				
			
ASSUME KERRY WON: 51.8% - 48.2%				
							
			HIGH	SOLID	NON-	HIGH	SOLID
	AVERAGE		BUSH	BUSH	PART. KERRY	KERRY
K/B	0.95		1.25	0.97	0.87	1.03	1.08
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.61%	39.10%	56.94%	59.84%	79.56%
Kpoll	49.52%		22.47%	38.21%	51.14%	60.93%	83.09%
WPE	4.55%		-7.71%	1.78%	11.59%	-2.19%	-7.07%
							
K/B	0.96		1.25	0.98	0.88	1.04	1.08
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.61%	39.00%	57.02%	59.82%	79.55%
Kpoll	49.92%		22.48%	38.41%	51.84%	61.06%	83.18%
WPE	3.77%		-7.73%	1.17%	10.36%	-2.48%	-7.27%

K/B	0.97		1.25	0.99	0.89	1.04	1.08
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.61%	38.90%	57.10%	59.80%	79.54%
Kpoll	50.30%		22.49%	38.61%	52.53%	61.18%	83.27%
WPE	2.99%		-7.76%	0.58%	9.15%	-2.76%	-7.46%
							
K/B	0.98		1.25	1.00	0.91	1.04	1.08
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.60%	38.81%	57.18%	59.78%	79.54%
Kpoll	50.69%		22.50%	38.81%	53.21%	61.30%	83.36%
WPE	2.23%		-7.79%	-0.01%	7.95%	-3.04%	-7.66%
							
K/B	0.99		1.25	1.01	0.92	1.05	1.09
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.60%	38.72%	57.26%	59.77%	79.53%
Kpoll	51.07%		22.51%	39.02%	53.88%	61.43%	83.45%
WPE	1.47%		-7.82%	-0.59%	6.75%	-3.32%	-7.85%
							

BASE CASE:
ALPHA = 1.0 

K/B	1.00		1.25	1.02	0.93	1.05	1.09
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.59%	38.64%	57.33%	59.75%	79.52%
Kpoll	51.44%		22.52%	39.22%	54.54%	61.55%	83.54%
WPE	0.72%		-7.84%	-1.16%	5.57%	-3.60%	-8.04%
							




K/B	1.01		1.25	1.03	0.95	1.06	1.09
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.59%	38.55%	57.40%	59.74%	79.52%
Kpoll	51.81%		22.53%	39.42%	55.20%	61.67%	83.63%
WPE	-0.03%		-7.87%	-1.73%	4.40%	-3.87%	-8.24%
							
K/B	1.02		1.25	1.04	0.96	1.06	1.09
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.59%	38.48%	57.46%	59.72%	79.51%
Kpoll	52.18%		22.54%	39.62%	55.84%	61.80%	83.72%
WPE	-0.77%		-7.90%	-2.28%	3.24%	-4.15%	-8.43%
							
K/B	1.03		1.25	1.05	0.98	1.06	1.10
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.58%	38.40%	57.52%	59.71%	79.50%
Kpoll	52.55%		22.55%	39.82%	56.48%	61.92%	83.81%
WPE	-1.50%		-7.93%	-2.83%	2.09%	-4.42%	-8.62%
							
K/B	1.04		1.26	1.06	0.99	1.07	1.10
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.58%	38.33%	57.58%	59.70%	79.50%
Kpoll	52.91%		22.56%	40.02%	57.11%	62.04%	83.90%
WPE	-2.22%		-7.95%	-3.38%	0.95%	-4.70%	-8.81%
							
K/B	1.05		1.26	1.07	1.00	1.07	1.10
Resp.	53.00%		40.00%	51.56%	51.76%	60.00%	60.00%
Kvote	51.80%		18.58%	38.26%	57.64%	59.68%	79.49%
Kpoll	53.27%		22.57%	40.22%	57.73%	62.17%	83.99%
WPE	-2.94%		-7.98%	-3.91%	-0.18%	-4.97%	-9.00%

.....................................................


		WPE SENSITIVITY TO ALPHA				

K/B	AVG	HB	B	BK	K	HK
 0.95 	4.55%	-7.71%	1.78%	11.59%	-2.19%	-7.07%
 0.96 	3.77%	-7.73%	1.17%	10.36%	-2.48%	-7.27%
 0.97 	2.99%	-7.76%	0.58%	9.15%	-2.76%	-7.46%
 0.98 	2.23%	-7.79%	-0.01%	7.95%	-3.04%	-7.66%
 0.99 	1.47%	-7.82%	-0.59%	6.75%	-3.32%	-7.85%

 1.00 	0.72%	-7.84%	-1.16%	5.57%	-3.60%	-8.04%

 1.01 	-0.03%	-7.87%	-1.73%	4.40%	-3.87%	-8.24%
 1.02 	-0.77%	-7.90%	-2.28%	3.24%	-4.15%	-8.43%
 1.03 	-1.50%	-7.93%	-2.83%	2.09%	-4.42%	-8.62%
 1.04 	-2.22%	-7.95%	-3.38%	0.95%	-4.70%	-8.81%
 1.05 	-2.94%	-7.98%	-3.91%	-0.18%	-4.97%	-9.00%

 
 

[TruthIsAll]

265. Some optimizer graphs

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x375366#376107


This was the run:

EXIT POLL RESPONSE OPTIMIZATION MODEL
6/7/05 10:04 PM

Objective:
Determine values of constrained variables required to derive
a target Kerry/Bush percentage split using aggregate exit
poll response data.


TARGET INPUT
Kerry 2-party vote 48.77%
Bush 2-party vote 51.23%
Wtd Avg Response 53.0%
Wtd Avg Alpha (K/B) 1.12

[TruthIsAll]

272. Update: slight fix to refuser calculation

EXIT POLL RESPONSE OPTIMIZATION MODEL							
6/10/05 7:34 AM							
							
Objective: Determine values of constrained variables required
to derive a target Kerry/Bush percentage split using aggregate
exit poll response data.							
							
Precinct Variables Range Constraints:							
1-Response: within (Min, Max); equals wtd. avg.							
2-Kerry win percentage: within (Min, Max).							
3-Alpha (K/B): within (Min, Max); equals wtd. avg							
4-WPE: within (Min, Max). 							
							
TARGET INPUT							
Kerry 2-party vote		52.50%					
Bush 2-party vote		47.50%					
							
Wtd Avg Response		53.0%					
Wtd Avg Alpha (K/B)		0.98					
							
RESPONSE INPUT CONSTRAINTS							
							
1250	Strong Bush		Strong Kerry			
Prcts	40	415	540	165	90		
							
KERRY WIN%							
Min	0%	20%	40%	60%	80%		
Max	20%	40%	60%	80%	100%		
							
RESPONSE							
Min	30%	30%	30%	30%	30%		
Max	70%	70%	70%	70%	70%		
							
ALPHA							
Min	0.10	0.10	0.10	0.10	0.10		
Max	10.00	10.00	10.00	10.00	10.00		
							
WPE							
E-M	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		
Min	-10.0%	-10.0%	-10.0%	-10.0%	-10.0%		
Max	10.0%	10.0%	10.0%	10.0%	10.0%		
							
OPTIMIZER OUTPUT SUMMARY							
							
	PERCENTAGE SPLIT		VOTE SPLIT (mm)				
PCT	POLL	VOTE	Diff	POLL	VOTE	Diff	
Kerry	51.77%	52.50%	0.73%	62.64	63.52	0.88	
Bush	48.23%	47.50%	-0.73%	58.36	57.47	-0.88	
							
Bush needed	46.65%	of refusers to match his vote					
							
PRECINCTS						Wtd Avg
Number	40	415	540	165	90		
Weights	3.2%	33.2%	43.2%	13.2%	7.2%		
							
Resp.	48.0%	49.9%	53.6%	57.4%	57.8%		53.0%
Dev	-5.0%	-3.1%	0.6%	4.4%	4.8%		0%
							
ALPHA							
K/B	1.20	1.02	0.92	1.06	0.96		0.98
Dev	22.3%	3.7%	-6.6%	7.9%	-1.7%		0%
							
2-PARTY VOTE 							
Kerry	20.0%	40.0%	55.6%	62.3%	87.8%		52.50%
Bush	80.0%	60.0%	44.4%	37.7%	12.2%		47.50%
Votes 							
(mm)	3.87	40.17	52.27	15.97	8.71		121.00
Kerry 	0.77	16.07	29.08	9.95	7.65		63.52
Bush	3.10	24.10	23.19	6.02	1.06		57.47
Diff	-2.32	-8.03	5.89	3.92	6.59		6.05
							
EXIT POLL							
Kerry	22.0%	40.3%	53.3%	64.1%	86.2%		51.77%
Bush	78.0%	59.7%	46.7%	35.9%	13.8%		48.23%
Votes 							
(mm)	3.87	40.17	52.27	15.97	8.71		121.00
Kerry 	0.85	16.20	27.85	10.23	7.51		62.64
Bush	3.02	23.97	24.43	5.74	1.20		58.36
Diff	-2.17	-7.77	3.42	4.49	6.32		4.29
							
WPE	-4.0%	-0.7%	4.7%	-3.6%	3.2%		1.46%
E-M	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		-6.77%
Diff	6.0%	5.4%	13.2%	2.3%	2.9%		8.2%
							
							
							
OPTIMIZATION MODEL							

Categ.	HighB	Bush	Even	Kerry	HighK		Total/Avg
Prcts	40	415	540	165	90		1250
							
VOTE							
Kerry	8	166	300	103	79		656
Pct	20.0%	40.0%	55.6%	62.3%	87.8%		52.50%
Bush	32	249	240	62	11		594
Pct	80.0%	60.0%	44.4%	37.7%	12.2%		47.50%
							
ALPHA							
K/B	1.20	1.02	0.92	1.06	0.96		0.98
AvgDev	22%	4%	-7%	8%	-2%		0%
							
RESPONDERS							
Total	19	207	289	95	52		663
Pct	48.0%	49.9%	53.6%	57.4%	57.8%		53.00%
							
Kerry	4	83	154	61	45		348
Pct	22.0%	40.3%	53.3%	64.1%	86.2%		51.77%
Bush	15	124	135	34	7		315
Pct	78.0%	59.7%	46.7%	35.9%	13.8%		48.23%
							
REFUSERS							
Total	21	208	251	70	38		587
Pct	52.0%	50.1%	46.4%	42.6%	42.2%		47.00%
							
Kerry	4	83	146	42	34		309
Pct	18.2%	39.7%	58.4%	59.9%	90.0%		53.35%
Bush	17	125	104	28	4		279
Pct	81.8%	60.3%	41.6%	40.1%	10.0%		46.65%
							
							
WPE							
Kv-Bv	-60.0%	-20.0%	11.3%	24.6%	75.7%		5.00%
Kp-Bp	-56.0%	-19.3%	6.5%	28.1%	72.5%		3.54%
							
WPE	-4.0%	-0.7%	4.7%	-3.6%	3.2%		1.46%
E-M WPE	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		-6.77%
							

[RonB]

283. A Note on "Maximal Standard Deviations" and their Use

Dear OTOH,

The "maximal standard deviation" (SQRT(0.5x0.5/n)) is the maximum possible sampling error that can occur when infering a true mean proportion from a sample of mean proportions.

These conservative overestimates of sampling standard deviation can thus show whether the variables (K and B) are STATISTICALLY equal across partisan precinct categories.

The fact that the actual data (assuming the O'Dell simulation has some relationship to the actual data) has much larger variance (and standard deviation) means that it contains discrepancies resulting from factors other then random error. This factors may be causing discrepancies because of measurement error in exit polling or because of vote miscount.

If these other factors (such as those listed in the E-M report: distance from precinct, interview rate, etc.) are acting more or less randomly throughout the sample - these large additional errors should cancel out so that the large variance will have zero bias - that is the mean (and median) of K and B should net out to what would result from random statistical sampling.

To the extent that these other factors produce mean (and median) K and B values that differ by more than the "random errror" difference (conservatively) calculated from the "maximal standard deviation", this shows bias in the action of these other factors that has to be explained.

There may be a reason why "neutral" exit polling factors happen to produce a significantly larger K in high Bush precincts - but this requires an explanation. It can't just be attributed to "large variance".

Large variances and measurement error are quite common in survey data - the key is to look at means and medians - and determine if they are "statistically different" (based on standard statistical assumptions) - this difference due to biased error (and not error that nets out to zero) is what has to be explained.

Best,

Ron

[OnTheOtherHand]

285. Maximal Standard Deviation

First a semantic note on "maximum possible sampling error," which will probably mean something different to most readers than it does to Ron. Ron is using "sampling error" here as basically synonymous with "standard deviation" or "standard error" -- and maximum because the formula assumes a p of 0.5, and higher or lower p values will lead to smaller standard errors. That's fine, although it doesn't mean that "the random error in any sample couldn't possibly be more than the MSD."

I think we can safely infer from the mean abs WPEs, or visual inspection of the scatterplot, or a cursory read of the E/M report in general, or first principles of survey research, that factors other than random sampling error enter into the exit poll results. But I ended where I should have begun. I don't know anyone who believes that an exit poll could have zero non-sampling error, even in the absence of fraud. Interviewing people at an election site is not like pulling ping pong balls out of a spinning barrel (or whatever it would take to get a pure simple random sample).

So, when we conduct statistical tests of the assumption that the non-sampling error is zero, and we reject that assumption, we have made no progress whatsoever toward assessing the probability of fraud. None whatsoever.

We can make no inferential progress without making some plausible assumptions about the possible magnitude of non-sampling error from sources other than fraud. In this case, we can infer either from E/M's report or from casual inspection of the scatterplot that there are substantial errors favoring Kerry as well as Bush, implying either that "innocent" non-sampling error is substantial or that there was considerable fraud in both directions. I can imagine many ways of trying to judge the plausible magnitude of innocent non-sampling error. Assuming that it should equal zero will not do.

Elizabeth Liddle initiated the convention, continued in several other simulations including some in the USCV working paper, of allowing K and B response rates to vary independently as a way of modeling innocent non-sampling error. It turns out that innocent non-sampling error can account for the observed means, medians, and variances, but of course that doesn't prove that it does. It may still be possible to find warrant to dismiss that null. But first one would have to take it seriously.

[OnTheOtherHand]

286. oh, hang on, you are modeling B and K...

(Now edited twice, the second time to tinker with terminology.)

So your argument, if I now understand it rightly, goes something like this:

Suppose that K (the completion rate* for Kerry voters) in every precinct is either 0% or 100%, which is an extreme assumption -- in fact, suppose that it is assigned to one condition or the other by coin flip. Even under that extreme assumption, the standard error of the mean of K across 40 precincts should only be about 8%. So to encounter a mean K of, say, 70% or more, as in the O'Dell simulations, is implausible.

* The E/M report gives "completion rates," "refusal rates," and "miss rates" for various sorts of precincts. Maybe because we spend so much time talking about "responders," we often use the phrase "response rates." I am trying to reform. BTW the reported completion rates probably aren't accurate, but we can at least use them as a rough baseline.

This argument actually fails on sampling error grounds, not (or as well as) non-sampling error grounds. The values of K that you would reverse-engineer from observed exit poll results are not comparable with actual completion rates.

For instance, suppose that a precinct is truly 80% Bush, but that the exit poll sample obtains 11 Bush and 9 Kerry respondents. This would be expected to occur by chance in perhaps 1% of precincts even if the true Kerry and Bush completion rates were equal (because the interviewer randomly approached more Kerry than Bush voters). Since we would "expect" to obtain 16 Bush and 4 Kerry voters, it follows that the relative completion ratio K/B is about 3.27. If I'm not confused yet again, I believe that assuming an overall completion rate of 56%, your formula would reckon this as completion rates of B = 39% or so, and K = 126% or so. Which is, of course, impossible, and yet could well occur through sampling error alone, and even with equal underlying completion rates.

I cringe to think what precinct-level values of K you could obtain by mechanically applying your formulas to observed results with both sampling and non-sampling error. Although, as far as I know, you have never actually calculated values of K for individual precincts, I don't see how a standard error of the mean of K could be interpreted without postulating values of K for individual precincts.

You may feel that "in the aggregate" this path strewn with logical impossibilities is nonetheless valid. I'm not sure what heuristic value it may possess, if we had no other way to tackle the variance problem. But we know from the O'Dell simulation and from my simulations that allowing the underlying, unobserved completion rates K and B to vary in individual precincts with a standard deviation of 10% per precinct, and adding sampling error, can produce results for the high-Bush precincts close to the observed ones. One such run of my simulation produced average underlying Bush and Kerry completion rates of 48% and 58%, reflecting a target alpha of 1.15 plus modest random variation -- plus precinct-level sampling error that presumably produced much wider swings in the B and K values your formula would calculate.

[RonB]

357. Maximal Standard Deviations - more clarification

OTOH (or Mark L?),

My math is fine. My equations for K and B are mathematically equivalent to those used by Liddle (alpha=K/B, Mitofsky (WPE_index=alpha), and O'Dell (alpha). I would encourage you to check this before claiming that my math is faulty. BTW, if my math did have an error, the way to show that is to point out the error rather than making vague general statements to that effect!

The R's calculated on Table 4 of Appendix F of the May 21 USCV paper, are based on K=.56 and B=.50 and the impossible neagtive and over 100% values (for high Kerry precincts) show that this E-=M hypothetical (as a constant value) cannot explain these data.
R is just backed out of the equations for K, B, and alpha.


Regarding the maximal standard deviations. An even more precise calculation is as follows:

For each precinct, for a perfect random sample K=B=R where is R is the precinct overall response rate. So each each Kerry or Bush voter has an R chance of completing the exit poll survey. This implies that the distribution of K for each precinct has mean=R and variance equal to: R(1-R)/nk where nk is the number of Kerry voters in the precinct(and similarly for B). Now we know that R will vary from precinct to precinct (though its average across precincts is .52-.56) so we conservatively set it at .5 to maximize precinct variance.

Since precinct samples are independent, the variance of the sum of these precinct specific random variables will be:

.25( 1/n_1k_1 + 1/n_2k_2 + .... + 1/n_40k_40)

where n_1,...,n_40 are the sizes of of the precincts for example for high Bush precincts 1 through 40 (N=40), and k_1,....,k_40 are their respective reported Kerry vote shares.

Now assuming:

(**) 1/n_1k_1 + 1/n_2k_2 + .... + 1/n_40k_40 <= 1

(*) for example if: n_ik_i >= 40 for i = 1 ...40


the variance of the sum will be less or equal to .25, and the standard deviation of the average of these proportions will thus be less or equal than: SQRT(.25/40)

One can see that because of (*), the SD of the average may be considerably less than SQRT(.25/N) when n_ik_i is >> than N, i.e. when N is small and k is large (for the high Kerry precincts) and probably for the the high Bush precincts (with at least 40 Kerry voters) as well. Especially in the high Kerry precincts this "Maximal SD" will probably be a considerable over estimate.

As the number of precincts in the quintile go up the value of
(**) may approach or even exceed 1 in some cases, even if all the precincts are quite large (n_i > N)

As the significant differences in K are between the high Bush precincts and all the others, I think there is a high probability that the levels of significance detected by this analysis are probably over estimates as precincts typically (as I recall) have many hundreds of voters - and particular the comparison between K's for high Bush and high Kerry precincts will be highly significant because of the likely smaller SD for high Kerry precincts.

Best,

Ron

[OnTheOtherHand]

360. faulty premises

All your work assumes, among other things, that B and K must be between 0 and 1, or 0% and 100%. However, as I demonstrated, a K (or B) calculated retrospectively from your formulas can be greater than 100%. (The actual response rates aren't observable.) So, there is no proof that your Maximal Standard Deviations are maximal. Nor does there seem to be much point in arguing about the distribution of your calculated B and K, since they obviously will not always equal the actual response rates.

I'm not even sure what you think your own results mean. Since O'Dell's mean abs WPEs are somewhat smaller than those reported by E/M and represented in the scatterplots, I don't see where you will get the leverage to argue that his variances are implausibly large. Insisting on using the "Maximal Standard Deviation," when it is obviously less than the actual standard deviation, will not work.

Table 4 of Appendix F poses similar conundrums. What exactly does it mean, or could it mean, to argue that mean and median WPEs close to 0 are "mathematically impossible" given an alpha of 1.12, regardless of the numbers of precincts and respondents? If it's just that the calculated mean alpha for the high-Kerry precincts won't equal 1.12, everyone concedes that at the outset. But if you want to argue for a statistically significant difference, then you need to introduce some variance into the model.

[tommcintyre]

361. "faulty premises", or "faulty qualifications on your part?

You're an Assistant Professor in Political Studies. Ron Baiman is listed as a statistician on the web (Google him, it comes up #1 on the web). And Kathy is a mathematician, right?

And Kathy says:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#376900
"People may want to know that our main two detractors, <you> and Bruce O'Dell have not contributed any mathematics to any of our papers to my knowledge..."

"Perhaps <you> and Bruce may just have a little difficulty understanding the math that Ron and I have done, and are emotionally attached to a position that is not supported by the facts..."

So, can you see why someone would have problems accepting your position over either Ron's and Kathy's?

[OnTheOtherHand]

364. not so much

With due respect to Ron, the fact that he is "listed as a statistician on the web" has little if any evidential force. Beyond that, I think that more balanced research yields a different picture, but I would encourage people to decide based on the arguments.

To further elucidate that point, and to respond to Dopp's comments: The willingness to resort to ad hominem argumentation -- nay, the insistence on doing so -- is not ordinarily understood by others as proof that one is winning the argument.

[tommcintyre]

367. Yes, very much

With due respect to Ron, the fact that he is "listed as a statistician on the web" has little if any evidential force. <"evidential force"??? We're talking about CREDIBILITY force! And, the fact that it's freepress.org (home of Bob Fitrakis) who lists him as such, actually carries a lot of weight around here (and should).>

Beyond that, I think that more balanced research yields a different picture, but I would encourage people to decide based on the arguments. <That would be a good idea if most people could UNDERSTAND the arguments. But, as the wikipedia quote I gave you earlier in this thread says: we must look to more authoritative/credible (demonstrated by qualifications) sources in such highly technical subjects. And, clearly, that is Ron and Kathy (mathematician).>

To further elucidate that point, and to respond to Dopp's comments: The willingness to resort to ad hominem argumentation... < Actually, Kathy is VERY SPECIFIC on the problems she has with you two. It seems when all else fails, you falsely claim an "ad hominem" attack has been made on you (and your cohorts). In reality, this is a fallacious argument by YOU, where you feign being the "victim".>

[mgr]

376. Qualifications:

Goggle based on Sunshinekathy's blowing OTOH's cover (I'm sure I commiting the ultimate faux pas--but this was available to Tom as well--a bit of bias I think):

"name"
Assistant Professor of Political Studies; Co-director of Environmental Studies Program

Program(s): Political Studies, Environmental Studies, American Studies, Studies in Race and Ethnicity (Affiliate)

A.B., Brown University; M.A., M.Phil., Ph.D., Columbia University. Jacob K. Javits Fellow (1992–96). Adjunct professor, Political Science Department and School of International and Public Affairs, Columbia University. Research interests: public opinion and public deliberation, American political institutions, quantitative methods, environmental politics. Contributor to Research in Micropolitics; Public Opinion; What Is America, and What Do We Want It To Be?: Foreign Policy and Domestic Priorities; The Role of the United States in a Changing World: Choices for the 21st Century. (2002– )

Looks to me as if he's well enough qualified--dead center. What part of 'quantitative methods' do you not understand?

Mike

[tommcintyre]

381. Hey! Nice puff piece mgr.

So, buried there in the middle is 'quantitative methods'. So, how many people can put THAT in their resume, do you think? And how many of them are NOT anywhere near on the same level as a "statistician" or "mathematician" regarding qualifications to analyze the subject at hand?

Bad try mgr. So much for the "obtuse" school of "reasoning". Honestly, I fear for those kids you teach.

[mgr]

398. I was not the one that posted the truncated version

I believe that you edited that same 'puff piece' up thread without the specifics indicating where in the epistomological universe OTOH exists. It's an honesty issue--you were dishonest. Dance around it all you want, and throw up all the dust you need but that is where it stands.

Mike

[Name removed]

403. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

405. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

406. Deleted message

Message removed by moderator. Click here to review the message board rules.

[mgr]

409. I have a singular problem with your defense.

How is it that I got for my first hit on Google the qualifications that I have, but yours are different. If anyone else were to do it would they get your results or mine? That is a scientific test--repeatability. You seem to be arguing for divine intervention.

"So, what I DID do was rearrange the order (to try to protect his anonymity - he wasn't fully "outed" at this time)"

I did not edit mine except to delete OTOH's name. I don't feel it necessary to provide the name of the institution where he works. You admit to tampering with the record--did you also edit the content? My version suggests that you had. You did not out him, but you just established where he teaches. Do you really value his anonymity? Or have you, made their life more difficult with that little bit of information?

A skunk is a member of the family Mustilidae, otherwise known as the weasel family.

You might also note that the statement:

"2) I'm putting you on permanent ignore. (Your the only person here at DU I've found it necessary to do this with.)....": violates the rules.

Have a good day,

Mike

[Name removed]

410. Deleted message

Message removed by moderator. Click here to review the message board rules.

[mgr]

411. That is a threat

On your PM you ask me to fess up or else you will go to the moderators. A little different from what you have here. You have my reply.

edited from n/t to clarify why it is a threat.

Mike

[tommcintyre]

412. That is really weak

C,mon, "be a man", "or face the music".

[mgr]

413. Care to show my reply n/t

[Name removed]

414. Deleted message

Message removed by moderator. Click here to review the message board rules.

[mgr]

416. Again, care to show my reply?n/t

[tommcintyre]

419. Care to "own up", and keep your word? n/t

[RonB]

377. Sampling Error vs Other Error

OTOH,

My work does not "assume" that K or B are between 0 and 1 any more than Liddle's or O'Dell's or Mitofsky's does. If you put in values for the other variables that drive K or B below 0 or above 1 - clearly you have an infeasible set of parameters - but mathematically this is perfectly possible just as an infeasible set of parameters for the high Kerry precincts (alpha=1.12) will result in impossible R values.

The O'Dell simulation SD's may very well be closer to those of the actual data, but they are not "random sampling SD" which is what I'm trying to estimate. Random sampling SD is a "confounding" SD which needs to be taken into account when looking at the mean and median results net of other sources of error.

This is exactly the same procedure that is done is estimating exit poll "margins of error" (and the same data - just different variables) - these are based on "sampling error" not the actual variance of the data.

The actual variance (as opposed to its net "bias" which shows up in all but the high Kerry precincts) is not what's important to establish statistical significance. When the difference in means results in a bias that is greater than what could be produced by random sampling - than it needs to be explained by other factors.

Regarding the calculations from means in the tables. These are estimates that (as it turns out are almost all within 1% of the O'Dell simulation estimates). They don't include variances necessary for statistical analysis and were not meant to.

My subsequent calcs do this. But these need to be calculated or estimated from sampling theory - not from "simulations" of, or indeed even from, the actual data.

Note for the record: I have done a lot of work and have taught many courses in applied statistics and was a math major and a math grad student (for a year), but my Ph.D. is in economics. I feel comfortable being called a "statistician" as long as its understood in the applied sense!

Best,

Ron

[OnTheOtherHand]

378. let me try to boil this down to two points

(1) The spreadsheet that spawned Table 4 indicates that not only is an alpha of 1.12 incompatible with the observed mean WPE of +0.3%, and median WPE of -0.4%, in high-Kerry precincts -- but an alpha of 1.001 is presumably incompatible because it would require the "overall response rate" to be -5.0% (based on the mean WPE) or +3.7% (based on the median WPE). Surely those response rates are highly implausible. Moreover, there is no value of alpha that will produce positive response rates based on both the mean WPE and the median WPE.

What do these results mean? Do they mean that any value of alpha is impossible? Is 1.001 somewhat less impossible than 1.12, and if so, how much less? And what does any of this have to do with the fraud debate?

(2) Are you surprised that the variability in error across precincts is larger than would be predicted for pure sampling error alone? Would you characterize this as a new result? What are the implications for the fraud debate?

[tommcintyre]

382. Why Ron is most credible on this subject:

<From his post you just replied to, in case you missed it:>
"Note for the record: I have done a lot of work and have taught many courses in applied statistics and was a math major and a math grad student (for a year), but my Ph.D. is in economics. I feel comfortable being called a "statistician" as long as its understood in the applied sense!"<my bolding>

I noticed you passed on commenting on that - why am I NOT surprised? ;)

[OnTheOtherHand]

399. funny, I thought that made _my_ point

I've done a lot of work in applied statistics, most of it on survey datasets.

I've only taught it twice at the graduate level; how many do I need before I get to be judged on the quality of my arguments?

I was never a math major, but I did win the freshman math prize in college, so I've never really thought of it as a glaring vulnerability.

If Ron feels comfortable being called a statistician as long as it is understood in the applied sense, then I suppose I am, too.

But you might like to share some of your thoughts on this subject with folks on the other thread who are disparaging PhDs.

[tommcintyre]

407. I would come to you first for advice on, say, environmental politics

such as ANWR (and then I'd still get a "second opinion" ;) ). But, after reviewing ALL the qualifications, I'd have to put Ron and Kathy first for statistical analysis.

And, there are other considerations. Ever since early on, I have seen you (and the others in your "camp") shuckin' an' jivin' (obfuscating and evading) the tough questions (that I happen to know about) that frankly test your "reliability" on those (like the very "hard" math) that most of us don't. (I have pointed this out to you and the others, often.)

Frankly, it seems like I've seen you all using too much of the "political" skills (spinning), and not enough of the "science" ones (with unambiguous answers/explanations). It comes down to a trust factor, really. So, if Kathy questions whether you really understand the math, I tend to believe her.

In contrast, your USCV "protagonists" have given clear straight answers. For example, Ron has never claimed his position "proves" fraud (in contrast,Mitofsky has claimed rBr proof - among other very unethical things that I have pointed out earlier). And then ODell's first post attempted to imply USCV was trying to claim proof, which I wasn't surprised Ron refuted in his next post.

I could go on and on with other examples; but the end result is: "past is prologue". Taking this into consideration, I consider their overall credibility high, and "your group", low.

As an example of this, I will ask you once again. Have you listened to the Victoria Collier interview (votescam - re: Mitofsky's election-theft complicity)?

Between the three of you, we have been asked to read: two papers, and multiple posts. Is it too much to ask YOU (and Febble) to reciprocate a little? Just maybe your position that "Mitofsky has NEVER been involved in a repub election theft" just might waver?

How can you ask us to consider your position, when you won't consider ours?

[RonB]

415. Response to Your Two Points

OTOH,

Finally a specific point to which I can give a specific answer. (Now I see what your problem has been!)

1) The problem is that you're not changing r (=(K+B)/2) as you change alpha. Its kind of complicated to use the formula in the table when you change alpha. Easier to use equation 3 in Appendix B to derive R from w (=B-K) - then don't have to bother with changing anything else. The formula comes out as: R=(2w/E)b(1-b)=(2w/E)k(1-k). You can verify that this formula gives the same results as the one in the Table (with alpha+1 in the denominator!).

Then set w= 0.005, i.e. assume that partisan response rates are almost identical with a slight edge to Bush voters. You'll get R=47% from the mean WPR.

This will give a negative response rate for the calculation from the median, indicating that the same precinct (with same b and k values) can't be both the mean and median precinct for this quintile.

If you however set w=-0.005 (giving a slightly higher response to Bush voters) you'll get R=35% from the median.

And another more general but key question:

2) I'm not at all surprised by the "volatility in error" across precincts. We know that many factors (besides sampling error) will effect differential partisan response. The question for analysis is the extent to which the mean or median of these errors (that are larger than sampling error) is non-zero, and whether these biased errors can be explaned by the factors. Do the factors act in such a way as to produce (on average and on median) large over response by Kerry voters in high Bush precincts, and almost zero (net) partisan response differential in high Kerry precints? This is what has to be explained with a multifactor analysis. One key part of this is to take out error that might simply be due to sampling that can "confound" the dependent variable here. I've worked out a system for this - coming soon - when I get the time!

[OnTheOtherHand]

421. actually, I'm not sure that you do

1. "The problem is that you're not changing r (=(K+B)/2) as you change alpha."

Sorry, but this reasoning does not seem to suffice. The reported completion rate in the high-Kerry precincts is 53.0%, so for an alpha close to 1, both B and K should be close to 53.0%, and therefore r = (K+B)/2 should be close to 53%.

I reprogrammed the spreadsheet to depend on alpha and to yield an overall response rate of 53%. For instance, for alpha = 1.12, the high-Kerry response rates are B = 48.2%, K = 54.0% (which "averages out" to 53% since there are more Kerry voters). This slightly alters results for alpha some distance from 1, but the basic characteristics are the same. The formula very tightly fits a slight positive alpha corresponding to the slightly negative median WPE, and a slight negative alpha corresponding to the slightly positive mean WPE. It insists, so to speak, that even slightly different alphas yield impossible response rates.

For instance, consider these R values from medians:

alpha = 1.01 -> R = 37.1%
alpha = 1.02 -> R = 73.6%
alpha = 1.03 -> R = 109.4%
alpha = 1.04 -> R = 144.8%

Different ways of assigning K and B will lead to somewhat different values of R, but I don't see how you can overcome the hypersensitivity of the model, since your denominator equals a multiple of the mean or median WPE (very small in these precincts).

Need I belabor the point that I don't much trust a model that insists that alpha absolutely must be 1.01, or maybe 1.02, but surely not 1.03 -- unless, of course, we look at means instead of medians, in which case it must be very close to 0.99, and surely not positive? Computationally this may be true (at least for some definition of alpha across a range of precincts), but I don't think it sheds any light whatsoever on plausible ranges of underlying response bias.

2. "I'm not at all surprised by the 'volatility in error' across precincts."

OK, we agree on that, and we agree (very importantly) that multifactor analysis is necessary. But then your Maximal Standard Deviations don't seem to be doing very much work. Nor is it obvious that we should assume away the actual standard deviations for purposes of statistical inference.

Best!

[OnTheOtherHand]

422. whoops, I meant positive or negative bias (ln alpha)

so, to obtain the median WPE -0.4%:

alpha = 1.01 -> R = 37.1%
alpha = 1.02 -> R = 73.6%
alpha = 1.03 -> R = 109.4%
alpha = 1.04 -> R = 144.8%

all entail a slight bias favoring Kerry, and presumably a realistic value would be somewhere in the range 1.01 to 1.02.

or, to obtain the mean WPE +0.3%:

alpha = 0.995 -> R = 25.0%
alpha = 0.990 -> R = 50.3%
alpha = 0.985 -> R = 75.7%
alpha = 0.980 -> R = 101.4% (and so on)

But is it true that if the observed median and mean WPE are what they are, then the actual alpha must be about 1.015, or 0.990, or perhaps both? No, neither. (First we would have to specify what we mean by "the actual alpha" -- but as O'Dell's simulation demonstrates, the observed mean and median are consistent with a mean alpha of 1.22 across 90 high-Kerry precincts.)

[RonB]

423. You were right! - E-M WPE's inconsistent with alpha = 1.12 regardless of R

Dear OTOH,

Thanks for catching me on this!

I should have realized this from the start (as I actually pointed out to Liddle that R cancels out in her formula for alpha - and more simply using the USCV formulas for K and B and the fact that alpha=K/B)!

There is only one alpha that is compatible with any given WPE, for any value of R. The equation is: alpha = ((k-0.5E)/k)/((b+0.5E)/b)
In other words you can't arbitrarily set K and B if WPE is given. Once you've found the alpha that's compatible with the given WPE, only then can you set one of B (or K) to determine R and the other K (or B).


For the E-M WPE means these are (going from high Kerry to high Bush as in Table 4): 0.99, 1.15, 1.19, 1.14, 1.38, calculated from medians these are: 1.01, 1.14, 1.18, 1.14, 1.21. As you can, except for the high Kerry precincts they are all above 1.12, and for the extreme partisan precincts they are way off in opposite directions - and different form each other in with a high degree of statistical significance as I've calculated in a previous post.

So I didn't even need to show that no plausible R is compatible with alpha=1.12 for these WPE's, I could have just shown that the WPE's themselves are incompatible with the E-M alpha "hypothetical" of 1.12.

Regarding my most recent post. The equation works in terms of w = B-K,
which is not determined by WPE in the way that alpha is.

(Parenthetically, Table 4 was a "stand alone" calculation that is not related to formulas for K, B, K/B=alpha, etc. in the report - If your interested I can give all the exact derivation of Liddle's equation for alpha from USCV equations for K and B - with alpha=K/B. These are all correct, as are the formulas for w, and r. I was working on the "given E there are two degrees of freedom" theory, K and B; w and r; alpha, R; etc. - but for alpha there are no degrees of freedom once E is given!)


Mitofsky has since said that the idea of "constant mean alpha" was a "USCV fabrication" - so he's backed off - or clarified - that idea. But the E-M hypothetical of 1.12 is inconsistent with the data even as a "weighted average" alpha that varies across precinct categories, and in any case if alpha varies like this it needs to be explained by other non-sampling factors - see previous posts.

I wouldn't put much stock in O'Dell simulation outcomes (especially for extreme partisan precinct categories) as they assume spurious correlations between R and E (going from high Kerry to high Bush precincts) of: -0.64, -.30, 0.15, 0.47, and 0.9! They also include variance of K and B that inexplicably goes from 2 times to almost 5 times the sample standard deviations for the competitive precincts - i.e they have an unexplainable relationship to sample sizes. See my post on this.

Finally, (unlike the Dopp simulations) they are based on single k and b values for each quintile. With the Dopp simulations which are uniform over all (1%) k and b values for all the quintiles - we were able to show that even if k and b are set to minimize the differences between simulated and E-M values - these values values are incompatible with K=0.56 and B=0.5 hypothetical - no big surprise in lieu of all the other ways that we have shown this!


Conclusion: 1) Large and unexplained variations in alpha in E-M data for extreme partisan districts regardless of response rate assumptions - calculated from means or medians so its not just a result of outliers. 2) Table 4, Appendix E, is erroneous and needs to be deleted from "Working Paper" and replaced by the simple calculations of alpha from E above.

Thanks much!

Ron

[OnTheOtherHand]

424. y'know, I'm not an environmental politics expert for nothing...

I'm glad we finally agree about the table.

I wish I could agree about the irrelevance of O'Dell's simulation outcomes. I don't see how you can say (if you are saying) that mean alpha in the high-Kerry precincts simply must be very close to 1, when O'Dell has given a counterexample. You are not required to put much stock in his simulation, but I don't think the enumeration of correlations between R and E settles the issue. If overall participation is allowed to vary across precincts, it seems reasonable that error will covary with the participation rate of the dominant partisans. As for the variances, hey, four outliers are four outliers.

I would rather not say much about the Dopp simulations at this point, since I guess they are going to be revisited. I think it's fair for O'Dell to point out that he has pretty much matched the means and medians (and started to come close on the mean absolutes), while none of the results in the USCV paper comes close. It might be illuminating to report the response rates required in the Dopp simulation to match the observed results.

[RonB]

425. Final Points - as this is become unproductive again

OTOH,

We had our moment when you finally came up with a specific point - but now its back to the misunderstanding and spin I guess.

I'm not "saying anything" - the equation (based on means and medians)dictates that alpha is basically 1 in high Kerry districts.
(Note that the more reliable part (the non partisan precincts) of even O'Dell's simulations match these aggregate calculations of alpha by within 1%)

O'Dell's "counter example" is based on unexplained correlations and variances that have no empirical basis - and are designed to produce a flattened alpha - this is data fabrication, not analysis.

Its not "reasonable" for R to be correlated in any way with E , especially not in relation to partisanship. One thing that, I thought, E-M, Liddle, and USCV, all agree on is that the variation in R is very small and probably statistically negligible across partisan districts.

-0.64 and 0.9 correlations between R and E effectively render O'Dell's results for high Kerry and high Bush precincts useless. Likewise the huge variances that have no relation to sample size - add to the lack of credibility of this exercise.

Matching medians and means is not what Dopp was trying to do. Bruce did it by obviously contrived means - but at least this was the goal of his simulation. In any case using this as a measure of comparison (as O'Dell does in his paper) makes no sense at all.

Ron

[Melissa G]

296. OK OTOH, Bruce is gonna do some math here this weekend

But i am confused about what you are doing/saying.
(I never thought this would be what I'd use those stats classes for which I took ohhh so long ago)

Ron writes..
snip...
The fact that the actual data (assuming the O'Dell simulation has some relationship to the actual data) has much larger variance (and standard deviation) means that it contains discrepancies resulting from factors other then random error. This factors may be causing discrepancies because of measurement error in exit polling or because of vote miscount.
snip...
Large variances and measurement error are quite common in survey data - the key is to look at means and medians - and determine if they are "statistically different" (based on standard statistical assumptions) - this difference due to biased error (and not error that nets out to zero) is what has to be explained.

mg...
He pretty much explains it like my old stats texts.. repeating the final line for emphasis...'this difference due to biased error (and not error that nets out to zero) is what has to be explained.'

You write various versions of...
If the exit poll was wrong, then I would expect some of the weighted demographics to be wrong as well. There is no magic formula for extracting good results from bad data. I don't know whether the ones you have cited are demonstrably wrong. (I actually did some work to explain why the preponderance of recalled Bush voters vs. Gore voters didn't seem so strange. I haven't done this work for other criteria; I don't think I'm obligated to be more curious about every single argument than the people who make the arguments.) But even if they are, it would not settle whether the 'raw' exit poll was right or wrong.

mg sums up..
I hear Ron saying here is the problem, and here is the math that explains the problem. There is an irregularity from that math that needs to be explained. The irregularity has yet to be explained but it looks verrry peculiar.

OTOH, I hear your response to be not that there is some possible mathematical error that Ron et al are pursuing but that the data is just bad and incapable of giving an accurate answer to the
question of whether or not there even were statistical anomalies that we should be concerned with. Am I hearing you accurately?

If so, what motivated you to join USCV in the first place and contribute to the paper. If not, please clarify what math needs to be fixed and how you would do it.
Thanks,
Melissa

[OnTheOtherHand]

297. good questions

I haven't had my coffee yet, so we may have to try it twice. (And the conversation can ramify in lots of directions, anyway.)

Ron B. and I basically agree that it is reasonable to look at the observed distributions and consider whether they are statistically and substantively significantly different from distributions we might expect (some null hypothesis or hypotheses).

We agree that it is important to look at biased error, not just error that nets out to zero.

I think that the way he calculates K and B is fundamentally flawed (which is why it can generate apparent completion rates over 100% or, in fact, under 0%), and that infirms a lot of his argument in this thread. I would consider that a mathematical error, although I don't think there is anything wrong with his algebra.

I think that he underestimates how much variance -- much of it presumptively innocent, since it favors Kerry (maybe I'm sentimental, but I'm not willing to embrace widespread Kerry fraud) -- is in the observed data. This informs tests of statistical significance. Remember how t or z scores are calculated as a difference from the mean, divided by the standard deviation or standard error? If you systematically lowball the standard deviations/errors, you will exaggerate the t scores.

(How to fix that? I've basically adopted and adapted the Liddle approach, which is to simulate varying Kerry and Bush completion rates, draw "samples" from various precincts with varying levels of partisanship, see how the statistics from those samples vary across multiple runs, and compare that with the reported results. O'Dell took a similar approach, hand-tweaked some values to increase the variance, but is still below the E/M reported mean absolute WPEs. Dopp also took a similar approach, but since no one here has discussed it, I will leave it be.)

Ron focuses on four outliers in high-Bush precincts as suggestive of fraud; I see outliers in all sorts of places and in both directions. I'm not saying that those four outliers (or the others in both directions) couldn't be fraud, although I think we already have places to look for fraud. But I do say that substantively, you could assume fraud in those precincts, double it, extrapolate to the whole country, and still come nowhere near changing the election outcome. The fixation on "Bush strongholds" seems unhelpful.

One remaining interesting question is whether the data nonetheless hint at a wider pattern of vote embezzlement. I'm not sure that Ron has actually addressed that question. So far, I think the lack of correlation between partisanship and the Febble function is a real problem for scenarios of massive vote embezzlement. But I agree with Ron that at a minimum, we need more multivariate analysis -- and I would add that we need more sharp thinking about what these scenarios might look like "on the ground" and how they would show up in the data.

(Another question is whether the whole idea of widespread response bias makes any sense. As far as I can tell, most public opinion researchers think that differential response does explain the exit poll discrepancies. But even if one assumes that that is true, there are still important questions that need to be sorted out.)

I joined USCV because I thought (and think) the work is important for the same reasons as Bruce -- but that isn't quite getting at your question. Would it have been intellectually correct to read E/M's January report, look at the mean absolute WPEs, and say, "Beam me up, there's no usable data here"? or to have bailed out before ever signing on? I sort of wish I had, in retrospect, but actually I don't think that's correct. Since then, for instance, we have explored facially plausible models of massive vote embezzlement that could be supported even by very noisy data (but apparently aren't -- but there could be other models). I have no way of anticipating what we might discover next. Even if we don't find massive vote embezzlement in 2004, the work may help us detect it or prevent it in future elections, and that is exactly USCV's mission as I understand it.

[Melissa G]

299. Amazing, it took us a while but we communicated successfully!

Thanks for this. I have more questions but others may get to them before me. (Housecleaning for a party and my already DU weary family will mount a mutiny if I get off task)
Best,
Melissa

[Melissa G]

304. OK ...Quick DU hit then back to the grind of my other life....

It seems as if we are all agreeing until P4 of your opus above where you say....
I think that the way he calculates K and B is fundamentally flawed (which is why it can generate apparent completion rates over 100% or, in fact, under 0%), and that infirms a lot of his argument in this thread. I would consider that a mathematical error, although I don't think there is anything wrong with his algebra.

mg...
So I hear you saying that you disagree with the way Ron has framed his mathematical arguments and that you believe he is asking the wrong questions.

You next say...
I think that he underestimates how much variance -- much of it presumptively innocent, since it favors Kerry (maybe I'm sentimental, but I'm not willing to embrace widespread Kerry fraud) -- is in the observed data. This informs tests of statistical significance. Remember how t or z scores are calculated as a difference from the mean, divided by the standard deviation or standard error? If you systematically lowball the standard deviations/errors, you will exaggerate the t scores.

mg
This is a place where I think there is much miscommunication..I think this may actually be what others are trying to tell you is more evidence of why they think there is fraud. I may be wrong and as we said before... concepts but not detailed math discussions with me... I actually do care what you think more than what AAPOR thinks because I will settle for dialog with one intelligent person at a time until we hit critical mass, which I think is coming soon. (Also we could could have a long discussion about all the things too many older, white, American men- especially when they get together in a group- get Wrong. :eyes: Tell me that is not what that AAPOR room looks like.)

If I have this correct. Others, jump in here please.. I am cringing because I hear you saying that your fix for this is modified rBr... Tell me I"m wrong..pretty please...If you really are doing this how do you answer, among oh so many others, our friend Bill Bored...
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=375366#375499

I agree with you that the fixation on **** strong holds is likely not a good idea. From spreadsheet work I remember done here earlier we suspect vote padding in Kerry strongholds to be significant in creating the false **** majority.
I'm glad you are here even if I cringe at some of your constructs because dialog with reasonable people who disagree is what furthers knowledge and echo chambers are less productive and a lot more boring.
I also appreciate the improvement in your writing style.
:toast:

[Melissa G]

255. My favorite paragraphs....

These findings show that (high Bush) precinct partisanship is strongly linked to a (calculated) unexplained increase in Kerry voter exit poll response. A linear correlation between “alpha” (Ln(K/B)) and precinct partisanship across the entire data set is irrelevant to this finding (see previous posts).
snip

Given the massive on-the-ground evidence of voting irregularities (circumstantial and other – including recent evidence of code in Diebold memory cards specifically allowing for vote corruption – see http://www.scoop.co.nz/stories/HL0307/S00064.htm ), it is reasonable to conclude that vote corruption is a viable alternative hypothesis for explaining these discrepancies.

[TruthIsAll]

269. Ron, this is the latest version of the model for 1.16 wtd alpha.

EXIT POLL RESPONSE OPTIMIZATION MODEL							
6/10/05 6:55 AM							
							
Objective: Determine values of constrained variables required
to derive a target Kerry/Bush percentage split using aggregate
exit poll response data.							
							
Precinct Variables Range Constraints:							
1-Response: within (Min, Max); equals wtd. avg.							
2-Kerry win percentage: within (Min, Max).							
3-Alpha (K/B): within (Min, Max); equals wtd. avg.							
4-WPE within (Min, Max). 							
							
TARGET INPUT							
Kerry 2-party vote		48.77%					
Bush 2-party vote		51.23%					
							
Wtd Avg Response		53.0%					
Wtd Avg Alpha (K/B)		1.16					
							
RESPONSE INPUT CONSTRAINTS							
							
1250	Strong Bush			Strong Kerry			
Prcts	40	415	540	165	90		
							
KERRY WIN%							
Min	0%	20%	40%	60%	80%		
Max	20%	40%	60%	80%	100%		
							
RESPONSE							
Min	30%	30%	30%	30%	30%		
Max	70%	70%	70%	70%	70%		
							
ALPHA							
Min	0.10	0.10	0.10	0.10	0.10		
Max	10.00	10.00	10.00	10.00	10.00		
							
WPE							
E-M	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		
Min	-10.0%	-10.0%	-10.0%	-10.0%	-10.0%		
Max	10.0%	10.0%	10.0%	10.0%	10.0%		
							
OPTIMIZER OUTPUT SUMMARY							
							
	PERCENTAGE SPLIT		VOTE SPLIT (mm)				
PCT	POLL	VOTE	Diff	POLL	VOTE	Diff	
Kerry	52.32%	48.77%	-3.55%	63.31	59.01	-4.30	
Bush	47.68%	51.23%	3.55%	57.69	61.99	4.30	
							
Bush needed	55.23%	of refusers to match his vote					
							
PRECINCTS						Wtd Avg
Number	40	415	540	165	90		
Weights	3.2%	33.2%	43.2%	13.2%	7.2%		
							
Resp.	48.0%	49.9%	53.6%	57.4%	57.8%		53.0%
Dev	-5.0%	-3.1%	0.6%	4.4%	4.8%		0%
							
ALPHA							
K/B	1.22	1.20	1.16	1.13	1.00		1.16
Dev	4.9%	3.7%	-0.2%	-2.5%	-13.4%		0%
							
2-PARTY VOTE 							
Kerry	20.0%	36.9%	50.0%	60.6%	86.9%		48.77%
Bush	80.0%	63.1%	50.0%	39.4%	13.1%		51.23%
Votes 							
(mm)	3.87	40.17	52.27	15.97	8.71		121.00
Kerry 	0.77	14.83	26.16	9.67	7.57		59.01
Bush	3.10	25.34	26.11	6.30	1.14		61.99
Diff	-2.32	-10.51	0.05	3.38	6.43		-2.98
							
EXIT POLL							
Kerry	22.2%	40.7%	54.0%	64.6%	87.1%		52.32%
Bush	77.8%	59.3%	46.0%	35.4%	12.9%		47.68%
Votes 							
(mm)	3.87	40.17	52.27	15.97	8.71		121.00
Kerry 	0.86	16.33	28.22	10.31	7.59		63.31
Bush	3.01	23.84	24.05	5.66	1.12		57.69
Diff	-2.16	-7.50	4.18	4.65	6.46		5.63
							
WPE	-4.3%	-7.5%	-7.9%	-8.0%	-0.4%		-7.11%
E-M	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		-6.77%
Diff	5.7%	-1.4%	0.6%	-2.1%	-0.7%		-0.3%
							
							
							
OPTIMIZATION MODEL							

Categ.	HighB	Bush	Even	Kerry	HighK		Total/Avg
Prcts	40	415	540	165	90		1250
Kerry	20%	37%	50%	61%	87%		48.77%
							
ALPHA							
K/B	1.22	1.20	1.16	1.13	1.00		1.16
AvgDev	5%	4%	0%	-2%	-13%		0%
							
RESPONDERS							
Total	19	207	289	95	52		663
Pct	48.0%	49.9%	53.6%	57.4%	57.8%		53.00%
							
Kerry	4	84	156	61	45		351
Pct	22.2%	40.7%	54.0%	64.6%	87.1%		52.32%
Bush	15	123	133	34	7		311
Pct	77.8%	59.3%	46.0%	35.4%	12.9%		47.68%
							
REFUSERS							
Total	21	208	251	70	38		587
Pct	52.0%	50.1%	46.4%	42.6%	42.2%		47.00%
							
Kerry	4	69	114	39	33		258
Pct	18.0%	33.2%	45.5%	55.2%	86.7%		44.77%
Bush	17	139	137	31	5		329
Pct	82.0%	66.8%	54.5%	44.8%	13.3%		55.23%
							
VOTE							
Kerry	8	153	270	100	78		610
Pct	20.0%	36.9%	50.0%	60.6%	86.9%		48.77%
Bush	32	262	270	65	12		640
Pct	80.0%	63.1%	50.0%	39.4%	13.1%		51.23%
							
WPE							
Kv-Bv	-60.0%	-26.2%	0.1%	21.1%	73.8%		-2.46%
Kp-Bp	-55.7%	-18.7%	8.0%	29.1%	74.2%		4.65%
							
WPE	-4.3%	-7.5%	-7.9%	-8.0%	-0.4%		-7.11%
E-M WPE	-10.0%	-6.1%	-8.5%	-5.9%	0.3%		-6.77%

[RonB]

284. TIA - please send me you model

Dear TIA,

Please send me your model - or model output, in Excel if you can.

Thanks!

Ron

[TruthIsAll]

300. Ron, what is your email?

tia

[RonB]

350. My Email

Dear TIA,

My email is: [e-mail]

Thanks!

Ron

[Bruce ODell]

257. An appeal to reason - with a detailed response this weekend

Due to recent family and professional obligations, I really have not had the time to compose a detailed response to Kathy's posting, nor to Ron's critique of my paper - for the moment. Since in some quarters absence of a response is treated as capitulation, however, I wanted to make my intentions clear: I do plan to post my detailed responses to their recent critiques this weekend.

In the meantime, I'll simply note here that as one of the founders of US Count Votes, I strongly believe in freedom of thought and inquiry and open debate, objectively grounded on data and the interpretation of data - free from unwarranted speculation about personal motivations, from appeals to emotion and from personal attacks.

I would ask all my critics in this forum to hold themselves to the same standard - if they can.

Robust debate is never "destructive" in pursuit of the truth! A "volunteer scientific research project" is successful when differing opinions are expressed, not suppressed.

I believe in these principles very strongly, and so I vigorously reject Kathy's suggestion I resign from the organization I helped found. The fact that I disagree with a paper published by a minority of USCV contributors is hardly grounds for expulsion from a research project - in fact instead it seems to me quite in keeping with the charter of our organization.

I will retract my paper - cheerfully - if and when proven wrong. I'll have more to say on that topic this weekend.

[autorank]

261. Please take is SOMEWHERE ELSE. We've had enough.

I don't know how to say it any more succinctly. Enough, just stop it. This is internal stuff, just call it a day, walk away, have your fights somewhere else. Since the start of May, there has been nothing here but elaborate, detailed philosophic debate augmented by some of the nastiest swarming attacks by the less polite of our detractors (all aimed at TIA, btw) that I've seen on any forum. I'm not saying you're folks brought the the nastiness but the atmosphere was certainly created for conflict. This is relay too much. Please stop. Get your own server and board if you want to fight with each other.

And visit this thread. It will let you know how people feel about the internal debates of USCV, all aired in public and fought out here. It's simply not fair.

A thread you should visit on DU elections. This is the exact title:
FOR PETE'S SAKE STOP THE DEBATE BETWEEN O'DELL AND USCV!

And while you're at it, why not...

Contact the DNC and Give 'em Hell About NOT Acting on Election Fraud

NEW LEADERS FOR A NEW DEMOCRATIC PARTY

[TruthIsAll]

264. Are you familiar with the Exit Poll Response Optimizer yet?

As a systems modeling expert, you should be VERY interested in target goal-seeking, method of steepest descent, non-linear and linear optimization, Newton/Raphson method, genetic algorithms, resource allocation and constrained optimization.

There is more to all this than just Monte Carlo Simulation.

Are you?

Finally, it all comes down to simple algebra.
And working WITH the numbers, not against them.

Just for kicks, what do you code in?

[Bruce ODell]

352. The tool should match the problem...

TIA, you've listed a number of interesting problem solving and constraint-based optimization techniques, none of which are meaningful to a general audience, but which certainly sound impressive.

I started my professional career twenty-five years ago in applied research and development of artificial intelligence techniques for satellite imagery analysis, robotic vision and industrial inspection systems and seismic data processing, and I have continued my professional education ever since.

I am a long-time member of ACM SIGART (ACM Special Interest Group on Artificial Intelligence, http://www.acm.org/sigs/sigart/ ), SIGSOFT (Special Interest Group in Software Engineering, http://www.acm.org/sigsoft/) and ACM SIGMETRICS (Special Interest Group on Measurement and Evaluation, www.sigmetrics.org) and I am an affiliate partner of Casewise, an enterprise modeling software vendor ( www.casewise.com ) with, in my opinion, the best process simulation tool on the market.

You have no basis for questioning my familiarity with techniques of constraint-based problem-solving, mathematical modeling or simulation.

The approach I chose was different than the approach you chose. To me, your model - and in fact most of the techniques you list - are simply not the best tools to simulate exit polls or the accuracy of elections.

The problem domain actually calls for a simple and obvious simulation technique. Think of my approach as being "object-oriented".

My goal was to create 1250 instances of simulated precincts, each with a simulated rate of Kerry and Bush exit poll participation and simulated actual vote. By doing so, I can investigate what combination of simulated precinct-level response rates and actual vote totals generate individual WPE and participation rate statistics that accurately roll up to match E/M's aggregate numbers.

The "constraint satisfaction" I built into my object-oriented model is simple to describe. For each of the five partisanship categories, I created a set of individual precincts that collectively match E/M's aggregate Mean WPE, Median WPE and overall precinct participation numbers published in January, while simultaneously matching the variance profile shown in the Mitofsky scatterplots published at AAPOR. You may recall I described the iterative process I used to satisfy those constraints in both the USCV Working Paper (Appendix G) and in my paper - if you've actually read them. Because my simulation approach was straightforward and object-oriented, I manually provided all adjustments needed to re-run my simulator and adjust the output to satisfy all constraints. Of course, you're certainly welcome to use program logic or spreadsheet functions to accomplish the same objective if your approach is more complicated than mine.

I'll also note that your simulation, unlike mine, does not satisfy the additional constraint imposed by the variance in the E/M scatterplot of precinct-level data. Do you plan to correct this oversight anytime soon? I think if you do, to the extent that your sample set of 1250 precincts satisfies the same set of constraints that mine does, and that are also reflected in the reported E/M data, you will necessarily see the same results I saw. The math is, what the math is.

In order to satisfy the variance constraint, I modeled separate Bush and Kerry response rates with pseudo-random variation on a Gaussian distribution with an adjustable mean. While my first-cut model of variance in response rates was reasonably close to those reported by Mitofsky at AAPOR, I increased the variance to match as best as I could visually estimate. However, the adjustments I made to match the extreme variance in the Mitofsky scatterplot were not rigorous. Yet I'm confident that with enough time spent measuring the dots on an enlarged printout of Mitofsky's scatterplot, I could even more closely reproduce the E/M WPE and Bias Index results. (I even think there is enough information to very closely estimate the actual precinct-level vote percentages and participation rates in the E/M sample, and I plan to do that when I have the time.)

Constraint-satisfaction algorithms based on aggregate data seem to me to have a fundamental conceptual mismatch with the problem space being simulated. I chose my object-oriented technique not because it supported my own foregone conclusion about the results, but because it allowed me to simply and naturally model the essential aspects of the problem space in a way that could be calibrated against the actual reported data. Creating a set of precincts that matches the actual data as best I can and then seeing what additional analysis of that data shows seems to me the simplest and most straightforward method of simulation, as well as being the tool most ideally suited to the problem.

This is quite different from your approach, which appears to presume that there is no credible set of individual precinct data which conforms to the aggregate constraints of Mean WPE, Median WPE and overall participation. This seems like a strange line of investigation, given E/M has actually released enough data to reconstruct the statistical profile of the individual precinct data. To me your approach implies that you do not trust Mitofsky's individual precinct scatterplot, but do trust last January's aggregate numbers. I don't understand your reluctance to actually try to analyze, or even reproduce the E/M precinct-level scatterplots.

Why do you continue to try to explain E/M aggregate numbers by every means other than the obvious one of actually correlating them with the available E/M precinct level data?

When you do look at the actual precinct-level data published in Mitofsky's scatterplots, or try to calibrate your simulation model against it - if you ever do look, and I'm increasingly less hopeful - you'll understand that there are no simple, pat answers.

There were extremes of very negative and very positive WPE (and response bias) in all categories of precinct. Are you concerned about extremely large positive WPE values, as well as negative ones? Large positive WPE indicates exit poll problems that under-polled Kerry supporters, or alternatively, vote count corruption favoring Kerry. What exactly is your explanation for the scatterplot precincts with very large positive WPE values - poll error or pro-Kerry vote fraud, or a combination of the two?

Whatever happened in 2004 remains unexplained. Once again, I believe an unknown combination of fraud by both Kerry and Bush supporters, plus problems with the exit poll together are the most likely explanation. But no one that actually looks at the E/M precinct-level scatterplot can say that there was a statistically significant relationship between the parallel explanations for the exit poll discrepancy (poll response bias/vote count corruption) and partisanship. And a close look at the scatterplot shows there's a lot of extreme positive WPE to explain as well.

To try to create elaborate constraint-satisfaction simulations to find a precinct-level solution for a problem expressed as a series of aggregate constraints - when it is so easy to reproduce both Mitofsky’s aggregate and precinct-level results via a simple simulation of 1250 individual precincts, calibrated to the available data - I think shows a fundamental misunderstanding of the problem at hand.

(BTW, as to the language I code in: I code only for fun now, in any language I'm interested in learning. Professionally it's been about fifteen years since I moved past coding to something to me far more interesting and challenging: developing the design and architecture of very large-scale systems and ensuring they perform and scale as expected by my customers. I'm expected to calibrate my models against the real world - are you?)

[tommcintyre]

353. QUESTION: Do you have "difficulty understanding the math"?

You're in computer networking (specializing in bank security?). Ron Baiman is listed as a statistician on the web (Google him, it comes up #1 on the web). And Kathy is a mathematician, right?

And Kathy says (in the excerpt below):

"...Bruce may just have a little difficulty understanding the math that Ron and I have done, and are emotionally attached to a position that is not supported by the facts..."

So, with all due respect, can you see why someone would have problems accepting your postion over Ron's and Kathy's?

Also, isn't it reasonable to have questions about what is said below?

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#376900

Bruce said that:

QUOTE: Robust debate is never "destructive" in pursuit of the truth! A "volunteer scientific research project" is successful when differing opinions are expressed, not suppressed. ENDQUOTE

There is a huge difference between "robust debate" and dishonest misrepresentations of the facts.

Making false statements is always destructive in my opinion, and Bruce has yet to retract any of the false statements that he continues to make about our working paper, despite my repeatedly pointing out to him in emails that the mischaracterizations he has been making about our working paper were not supported by any sentence in our paper.

I have asked Bruce dozens of times to constructively find the exact sentence, or mathematical or logic error anywhere in our current working paper. He has never done so.

I informed Bruce repeatedly over a period of weeks that the mischaracterizations he made about our paper were not supported by any sentences in our paper that I knew of, and asked him to find the sentences in our paper that matched his characterization of it. He never did.

I have offered many many times via emails to work with Bruce if he would specify even one exact sentence in our paper that he did not agree with, yet he has never taken me up on my offer to listen to any constructive criticism he had of the paper.

I stayed up all night the night before the AAPOR conference trying to incorporate as many of the points of Bruce's last minute, paper that he insisted we substitute for the one all of us had been working on, trying to please him.

I can only guess what O'Dell's motivations are, but his actions have been and remain very destructive.

Distorting the truth is always destructive IMO and never has a good outcome and should be retracted.

People may want to know that our main two detractors, Mark Lindeman and Bruce O'Dell have not contributed any mathematics to any of our papers to my knowledge, although Bruce has contributed some programming work. Both Bruce and Mark seem to excel at writing and communicating.

Perhaps Mark and Bruce may just have a little difficulty understanding the math that Ron and I have done, and are emotionally attached to a position that is not supported by the facts without some pretty implausible assumptions about the size of exit poll variance that are not generally accepted.

[autorank]

355. Well, this looks like the final word. Time to call it a thread.

[OnTheOtherHand]

358. or something like that...

Tom, if you have actually read Bruce's and my posts, you know that we both have quant credentials. If Ron wants to play dueling credentials with me, I guess we can, but I decline to do it at your behest.

Someone's assertions that someone else may have trouble understanding their math doesn't cut much ice unless the math is right. Or, at least, there is an argument about the math.

For that matter, there's a lot in Kathy Dopp's post that doesn't cut much ice. Trying to set aside the personal stuff... hmm. (O'Dell has no specific criticisms -- that's just weird. "Not contributed any mathematics to any of our papers" -- well, that's shading things a bit fine.) Oh, yes, here's something about the size of exit poll variance. I pointed out elsewhere that the variance in O'Dell's simulation is actually less than the variances reported by E/M back in January. But as I remember the history of this argument, that probably isn't Dopp's objection.

Yup, 'rank is right, probably time to call it a thread.

[tommcintyre]

363. You'd make the "Artful Dodger" proud... well, at least for trying ;)

Tom, if you have actually read Bruce's and my posts, you know that we both have quant credentials. <You know, I've read all-too-many of your posts - IMO, they tend to have the same general themes of obfuscation and avoidance. ;) > If Ron wants to play dueling credentials with me, I guess we can, but I decline to do it at your behest.<See what I mean?>

Someone's assertions that someone else may have trouble understanding their math doesn't cut much ice unless <the person is more qualified than you, and, as she says, you haven't contributed any math. WHERE'S THE BEEF (substance)?>

Yada, yada, yada...

BTW: Have you decided - are you going to help us to either "blow up" (as you put it); or exonerate Mitofsky?

I'm still waiting for a CLEAR, SPECIFIC answer from you on this; but considering your "track record", I'm not holding my breath.

[OnTheOtherHand]

365. ahh

Yes, I am avoiding a direct comparison of credentials, because I believe it is irrelevant -- but no, I do not kneel before Dopp's masters degree or Baiman's first hit on Google.

Dopp has been known to say many things. I would as soon not get into a public spitting match with Dopp, especially if she relies on a proxy to do the spitting.

On the last, let me try a one-word response: no.

I will neither seek to "blow up" nor to "exonerate" Mitofsky. I will try to make the best use of Mitofsky's information, and to get more. When I disagree with Mitofsky, I will say so. As a matter of professional courtesy and simple human decency, I will observe that in my opinion, no compelling (or even very suggestive) case has been made that he has ever attempted to cover up anything to benefit Republicans. USCV's work has depended on Mitofsky's basic integrity. That could be wrong, but no fair-minded person should assume that it is. Thank you for this opportunity to clarify my position.

[tommcintyre]

368. "Ohh"

Yes, I am avoiding a direct comparison of credentials, because I believe it is irrelevant <as the wikipedia quote I gave you earlier in this thread says: we must look to more authoritative/credible (demonstrated by qualifications) sources in such highly technical subjects.> -- but no, I do not kneel before Dopp's masters degree <IN MATHEMATICS! - most relevant to the subject> or Baiman's first hit on Google.<...the fact that it's freepress.org (home of Bob Fitrakis) who lists him as such, actually carries a lot of weight around here (and should).>

Dopp has been known to say many things. I would as soon not get into a public spitting match with Dopp, especially if she relies on a proxy to do the spitting. <As usual, pretty assumptive on your part. Kathy certainly doesn't "rely" on any proxies. SHE made the original post; and I have independently (and voluntarily) chosen the quote her when relevant, because SHE IS ONE OF THE TWO MOST CREDIBLE SOURCES ON THE SUBJECT (Ron being the other one).

On the last, let me try a one-word response: no. <Finally, a straight answer! You've been "dancing around this" one ever since you first said I had no evidence a few threads ago. I repeatedly offered you the evidence, and you uniformly failed to even acknowledge the offer (as have the others who share your pro-Mitofsky/rBr viewpoint).>

"...no compelling (or even very suggestive) case has been made that he has ever attempted to cover up anything to benefit Republicans."
<WOW! What a statement (considering ALL the evidence). Did you even bother to listen to Victoria Collier's interview (at a bare minimum), as I suggested when I thought we were finally having an open and honest dialog about this (when I suggested even the possibility of exonerating him - if that is remotely possible). (To assist you, I'll list our conversation regarding this again here (since it's so hard to find in this increasingly ungainly thread). I am sincerely hoping you will finally take the time to listen to this interview (assuming you haven't already done so), and then please share your thoughts on this:>

<Literally, I've just stumbled on most of these. I haven't even seriously investigated him yet. This was part of a post originally addressed to Febble>

Regarding Mitofsky:
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x371726#373177
(I've pasted the most important parts here:)
"...you revealed some interesting things about Mitofsky:

1) He claimed rBr as fact in his original report. At MOST, this was a hypothesis, he knew that, EVEN his rBr "supporters" call it that. Plain and simple, legitimate professionals NEVER claim a hypothesis as a fact - by definition, it is simply a "guess", or "hunch" that must be investigated further. THIS ACTION ALONE IS A STRONG BREACH OF ETHICS. Thanks for pointing it out.

2) As you also pointed out, he also advanced rBr WITHOUT ANY SUPPORTING EVIDENCE. Again, this is something no ethical professional would dream of doing, in a formal report, as he did. They would wait until the investigation was reasonably done, and ONLY THEN publish their findings. Again, Thanks for pointing it out.

In your post above, you say:
"I know almost nothing about him..."
When you find the time, it would be well worth your while to learn more about him. Here's some info to get you started:

---------------------------
Febble, would you at least listen to this interview? I've read your paper twice, printed it out, and highlighted it and notated it, so this isn't too much to ask, is it? ;)

OTOH, will YOU listen to this interview? I doubt it.

3. I've listened to Victoria Collier's interview, and highly recommend it

http://www.edwardsdavid.com/BushVideos/infidelguy.com_Vote_Scam_a.mp3
Also, read the first eight chapters (three more than at the votescam site) of the book here:
http://www.constitution.org/vote/

<Info on when Mitofsky first "invented" rBr for Bush Sr. in 1992>
*<Verification here:>
http://www.votefraud.org/Archive/Write/newhampad.htm
"...early media reports stated that Bush and Buchanan were running neck and neck in the 1992 New Hampshire primary..."

<My original source:>
http://www.votefraud.org/News/2000/7/071800.html

"Thank you for this opportunity to clarify my position."

<You are ENTIRELY welcome!>

[autorank]

370. The Chernobyl thread lives on....

What are you doing on this issue besides sitting here and engaging in banter. Are you publishing someting on election fraud, working on a project, part of a movement. What do you do on this besides hang out here and npost?

[TruthIsAll]

354. Thanks for responding, Bruce. Now its my turn.

Thanks for the lengthy response. I will reply and be brief. Verbiage is not my strong suit.

You:TIA, you've listed a number of interesting problem solving and constraint-based optimization techniques, none of which are meaningful to a general audience, but which certainly sound impressive.

Me: They were meant for you, not for the general audience. I’m glad you think they sound impressive. But my question was simply if you were familiar with them. You have not responded to that question.

You:I started my professional career twenty-five years ago in applied research and development of artificial intelligence techniques for satellite imagery analysis, robotic vision and industrial inspection systems and seismic data processing, and I have continued my professional education ever since.

I am a long-time member of ACM SIGART (ACM Special Interest Group on Artificial Intelligence, http://www.acm.org/sigs/sigart / ), SIGSOFT (Special Interest Group in Software Engineering, http://www.acm.org/sigsoft /) and ACM SIGMETRICS (Special Interest Group on Measurement and Evaluation, www.sigmetrics.org ) and I am an affiliate partner of Casewise, an enterprise modeling software vendor ( www.casewise.com ) with, in my opinion, the best process simulation tool on the market.

You have no basis for questioning my familiarity with techniques of constraint-based problem-solving, mathematical modeling or simulation.

Me: Why not? It's germane to the problem at hand.

You: The approach I chose was different than the approach you chose. To me, your model - and in fact most of the techniques you list - are simply not the best tools to simulate exit polls or the accuracy of elections.

Me: And what brings you to that conclusion? Is it your specific knowledge of these techniques?


You: The problem domain actually calls for a simple and obvious simulation technique. Think of my approach as being "object-oriented". My goal was to create 1250 instances of simulated precincts, each with a simulated rate of Kerry and Bush exit poll participation and simulated actual vote. By doing so, I can investigate what combination of simulated precinct-level response rates and actual vote totals generate individual WPE and participation rate statistics that accurately roll up to match E/M's aggregate numbers.

Me: Why roll-anything up? Why not use an algorithm-based search to determine feasibility, subject to the given problem constraints and see what you come up with?

You: The "constraint satisfaction" I built into my object-oriented model is simple to describe. For each of the five partisanship categories, I created a set of individual precincts that collectively match E/M's aggregate Mean WPE, Median WPE and overall precinct participation numbers published in January, while simultaneously matching the variance profile shown in the Mitofsky scatterplots published at AAPOR. You may recall I described the iterative process I used to satisfy those constraints in both the USCV Working Paper (Appendix G) and in my paper - if you've actually read them. Because my simulation approach was straightforward and object-oriented, I manually provided all adjustments needed to re-run my simulator and adjust the output to satisfy all constraints. Of course, you're certainly welcome to use program logic or spreadsheet functions to accomplish the same objective if your approach is more complicated than mine.

Me: What were the criteria for applying these “manual” adjustments?
And how did you define your constraints? My approach is not more complicated than yours; it is far simpler. I’m lazy, I believe in letting the computer do all the work. I only specified the precinct range constraints. How did you specify your constraints?

You: I'll also note that your simulation, unlike mine, does not satisfy the additional constraint imposed by the variance in the E/M scatterplot of precinct-level data. Do you plan to correct this oversight anytime soon? I think if you do, to the extent that your sample set of 1250 precincts satisfies the same set of constraints that mine does, and that are also reflected in the reported E/M data, you will necessarily see the same results I saw. The math is, what the math is.

Me: First of all, let’s get our definitions straight. My approach is constrained optimal goal-seeking, not simulation. I define the problem to achieve an objective target, specify all the constraints, and see if the model returns a FEASIBLE solution which must satisfy ALL constraints.

You: In order to satisfy the variance constraint, I modeled separate Bush and Kerry response rates with pseudo-random variation on a Gaussian distribution with an adjustable mean. While my first-cut model of variance in response rates was reasonably close to those reported by Mitofsky at AAPOR, I increased the variance to match as best as I could visually estimate. However, the adjustments I made to match the extreme variance in the Mitofsky scatterplot were not rigorous.

Me: Why not?

You: Yet I'm confident that with enough time spent measuring the dots on an enlarged printout of Mitofsky's scatterplot, I could even more closely reproduce the E/M WPE and Bias Index results. (I even think there is enough information to very closely estimate the actual precinct-level vote percentages and participation rates in the E/M sample, and I plan to do that when I have the time.)

Me: YOU NEED MORE TIME TO MEASURE THE DOTS?

You: Constraint-satisfaction algorithms based on aggregate data seem to me to have a fundamental conceptual mismatch with the problem space being simulated. I chose my object-oriented technique not because it supported my own foregone conclusion about the results, but because it allowed me to simply and naturally model the essential aspects of the problem space in a way that could be calibrated against the actual reported data. Creating a set of precincts that matches the actual data as best I can and then seeing what additional analysis of that data shows seems to me the simplest and most straightforward method of simulation, as well as being the tool most ideally suited to the problem.

Me: I disagree. I have obtained the same alpha (1.15) that you did, without having to create an object-oriented set of precincts, or reading dots in scatter-plots. You are defending your simulation approach. There is no need to do that. Ron Baiman has used it. But to say that it is the BEST approach is quite presumptuous on your part.

You: This is quite different from your approach, which appears to presume that there is no credible set of individual precinct data which conforms to the aggregate constraints of Mean WPE, Median WPE and overall participation.

Me: I do not presume that. After all, the precinct data could have been mistabulated for all we know. In fact, later I will show you just how it could have been.

You: This seems like a strange line of investigation, given E/M has actually released enough data to reconstruct the statistical profile of the individual precinct data. To me your approach implies that you do not trust Mitofsky's individual precinct scatterplot, but do trust last January's aggregate numbers. I don't understand your reluctance to actually try to analyze, or even reproduce the E/M precinct-level scatterplots.

Me: And I do not understand why Mitofsky has not released the detailed precinct data.

You: Why do you continue to try to explain E/M aggregate numbers by every means other than the obvious one of actually correlating them with the available E/M precinct level data?

Me: That’s news to me. I never knew that the precinct data was made available.

You: When you do look at the actual precinct-level data published in Mitofsky's scatterplots, or try to calibrate your simulation model against it - if you ever do look, and I'm increasingly less hopeful - you'll understand that there are no simple, pat answers.

Me: You mean I have to actually LOOK at those scatter plots? We want the data man. Pure and simple.

You: There were extremes of very negative and very positive WPE (and response bias) in all categories of precinct. Are you concerned about extremely large positive WPE values, as well as negative ones? Large positive WPE indicates exit poll problems that under-polled Kerry supporters, or alternatively, vote count corruption favoring Kerry. What exactly is your explanation for the scatterplot precincts with very large positive WPE values - poll error or pro-Kerry vote fraud, or a combination of the two?

Me: Did you ever consider a Reluctant Kerry Responder? Or are you wedded to the Reluctant Bush Responder hypothesis?

You: Whatever happened in 2004 remains unexplained. Once again, I believe an unknown combination of fraud by both Kerry and Bush supporters, plus problems with the exit poll together are the most likely explanation.

Me: You are whitewashing the manifestly overwhelming documented evidence of Bush fraud (86 of 88 touch screens turned Kerry votes to Bush votes. You are attempting to obfuscate the facts that:
1)Exit polls are historically very accurate
2)The great preponderance of evidence on the ground in OH and elsewhere that virtually ALL anomalies and machine “glitches” favored Bush.

You: But no one that actually looks at the E/M precinct-level scatterplot can say that there was a statistically significant relationship between the parallel explanations for the exit poll discrepancy (poll response bias/vote count corruption) and partisanship. And a close look at the scatterplot shows there's a lot of extreme positive WPE to explain as well.

Me: So now you say it is a combination of polling error and Bush fraud AND Kerry fraud. Well one out of three hits and you’re batting .333.

You: To try to create elaborate constraint-satisfaction simulations to find a precinct-level solution for a problem expressed as a series of aggregate constraints - when it is so easy to reproduce both Mitofsky’s aggregate and precinct-level results via a simple simulation of 1250 individual precincts, calibrated to the available data - I think shows a fundamental misunderstanding of the problem at hand.

Me: I wrote my model which replicates the essential points that Ron Baiman has made in a matter of a few days WITHOUT that scatter plot you keep referring to. I think you have a fundamental misunderstanding of how weak your argument is. I will prove this to you by asking you, as I have asked others, to explain the following in plain English:

As I am sure you are aware, the National exit poll at 13047 respondents had Kerry winning by 51-48%. At the 13660 final, Bush was the 51-48% winner. Of course, how this could occur with just 613 additional responders is a mathematical impossibility. But let’s not dwell on that. Instead, I ask you to explain this VERY SPECIFIC demographic category weighting from the exit poll: How did you vote in 2000?

At the 13047 timeline, 41% said they voted for Bush, 39% for Gore. And Kerry won that poll. In the 13660 timeline the split was 43% Bush/37% Gore. But the 41%/39% was verifiably incorrect. In fact, it was the ONLY demographic which can be checked against actual historic numbers. Here is how we know that. We know that in 2000, Gore got 50.999 million votes, Bush 50.456. WE also know that approximately 3.5% of those voters died. That means that a MAXIMUM of 48.7 million Bush voters could have voted in 2004. Now 48.7/122.3 = 39.82%. That was Bush’s ABSOLUTE MAXIMUM.

So my question to you is this: How can anyone believe the final 13660 National Exit Poll, when we know that the adjustment in the weighting was made from an IMPOSSIBLE 41% to an even MORE OUTRAGEOUS 43%. THIS WEIGHTING WAS NECESSARY TO MATCH A CORRUPTED VOTE COUNT. THE 43% IS MATHEMATICALLY IMPOSSIBLE. THAT IS PROOF IN AND OF ITSELF THAT OVER 4 MILLION VOTES WERE SWITCHED FROM KERRY TO BUSH. AND THAT IS WHY THE EARLY EXIT POLLS WERE INDEED CORRECT, BECAUSE WE NOW KNOW THAT IT WAS THE FINAL 13660 THAT WAS BOGUS.

WHICH ONE DO YOU BELIEVE: THE 13660 OR THE 13047 TIMELINE ? WHAT IS YOUR EXPLANATION FOR THE CHANGE FROM 41% TO 43%? HOW DO YOU JUSTIFY IT? DO YOU CALL THAT SAMPLING ERROR? DO YOU CALL THAT EVIDENCE OF KERRY FRAUD? OR DO YOU CALL IT SOMETHING ELSE?

You: (BTW, as to the language I code in: I code only for fun now, in any language I'm interested in learning. Professionally it's been about fifteen years since I moved past coding to something to me far more interesting and challenging: developing the design and architecture of very large-scale systems and ensuring they perform and scale as expected by my customers. I'm expected to calibrate my models against the real world - are you?)

Me: That is very commendable. Now calibrate your model against the real-world numbers I have just given you.

Best
TIA

[eomer]

369. Bruce, help me understand what you are saying.

Bruce (or OTOH as Bruce's proxy), can you confirm my
understanding of your simulation?  I'm looking at your
USCV_ODell_Simulation2.xls spreadsheet.  In the "Summary
Charts" tab you have an area titled "Simulation
1".

Columns A & B of that area of the spreadsheet are as
follows:

Bush vote	Bias
   87	        16.4%
   67	         6.3%
   50	         8.3%
   33	         8.1%
   13	         2.1%

Columns A, G & H are as follows:

Bush vote   Mean Kerry Participation  Mean Bush Participation
   87	            70.4%	                54.0%
   67	            59.6%	                53.3%
   50	            56.3%	                48.0%
   33	            57.8%	                49.7%
   13	            53.6%	                51.5%

Can you confirm my understanding of the following points?

1) The response rates shown above are the rates required as
input into  your simulator in order to reproduce the overall
response rates, mean WPEs and median WPEs of the actual data.

2) You followed the iterative process described in Appendix G
of the USCV paper to come up with these response rates.

3) You wrote Appendix G of the USCV paper.

4) The response rates presented in the USCV paper (May 21
version) in the chart on page 10 were developed by you (these
are basically the same response rates shown above but in chart
form).

5) You agree with the part of the USCV paper that says that
the response rates shown above are the response rates you
would need as assumptions if you want to reproduce the actual
data without any other source of bias.


edit: improve wording for clarity

[OnTheOtherHand]

371. required participation rates

(edit to correct typo)

Eomer, I don't speak for Bruce on this issue, but I am happy to say what I think about it.

I haven't checked the appendices etc., but Bruce's simulation would indeed imply mean Kerry participation rates around 70% in the high-Bush precincts. This debate isn't so much about the descriptive stats, especially ones as subtle as "participation rates" seem to be.

Note, first of all, that this doesn't mean one has to enter a mean Kerry participation rate of 70% (or whatever) into the O'Dell simulator in order to approximate the reported E/M results. So, point (5) I think is wrong. One doesn't assume a Kerry response rate of 70% in order to obtain the O'Dell result; the 70.4% is a computed result, not an input assumption. I'm not sure what the input assumption was. (I can obtain similar results with an alpha of 1.15, or even a lower alpha with more variance -- although, as I note below, I have a somewhat different strategy.)

Second, it is important to realize that the O'Dell simulator doesn't model individual-level sampling error. This is important because, as I've pointed out, sampling error can produce widely ranging post hoc estimates of Kerry and Bush participation rates even if the actual completion rates are equal. In much of Baiman's work, AFAICS, any sampling variance is rendered as differential "response rates." A precinct exit poll result of "24% Kerry" when the official value is "18% Kerry" will appear as a much higher "response rate" for Kerry voters in that precinct, even if that difference could easily be attributed to sampling error alone, with no actual difference in completion rates between Bush and Kerry voters. (And of course it could be attributed to a combination of sampling error and differential completion rates.)

Note that it is perfectly possible to generate "mathematically impossible values" of K or B as calculated in the USCV working paper, with over 100% "response rates," in individual precincts from sampling error alone. (At least, I think so -- I haven't yet gotten Baiman to confirm or deny. It might be possible to finesse the K and B response rates as merely bizarre, but that wouldn't really alter the underlying point.) So, post hoc K and B are hardly even meaningful, as far as I can tell.

O'Dell's situation is somewhat different: he derives his estimated votes reported (which are fractional) from his participation rates, which do take on possible values in every precinct. His participation rates are prior to his estimated votes reported, not post hoc. Nevertheless, his model can only model (for instance) the distinction between "24% Kerry" in the exit poll and "18% Kerry" in the official returns as the outcome of a difference in participation rates. That is a reasonable conclusion, depending on how you understand "participation rates" -- Kerry voters apparently "participated" at a higher rate. But it doesn't necessarily imply that Kerry voters who were approached responded at a higher rate. So, O'Dell's participation rates don't give us direct leverage on "reluctant Bush responder" etc.

I believe it is because O'Dell's model doesn't model sampling bias that its variances run low: by default, it simply excludes extreme outliers that would result in extreme post hoc values of K and B (and, of course, suppresses variance overall).

My simulations, run in Python, model sampling error directly. As I reported in #286, one run of my simulation "produced average underlying Bush and Kerry completion rates of 48% and 58%, reflecting a target alpha of 1.15 plus modest random variation -- plus precinct-level sampling error that presumably produced much wider swings in the B and K values formula would calculate."

Obviously I'm not pounding the table and insisting that my simulation depicts what really happened. But I certainly am not convinced that the results from the 40 high-Bush precincts are statistically bizarre.

[eomer]

372. Regarding point 5...

Here's the part of Bruce's post #352 that I'm trying to understand:

The "constraint satisfaction" I built into my object-oriented model is simple to describe. For each of the five partisanship categories, I created a set of individual precincts that collectively match E/M's aggregate Mean WPE, Median WPE and overall precinct participation numbers published in January, while simultaneously matching the variance profile shown in the Mitofsky scatterplots published at AAPOR. You may recall I described the iterative process I used to satisfy those constraints in both the USCV Working Paper (Appendix G) and in my paper - if you've actually read them.

If I read that correctly, Bruce says that he used the process described in USCV Appendix G. One step of the process in Appendix G is as follows:

The simulation accepts an exit poll participation bias parameter for either Bush or
Kerry voters which moves the mean of the participation curve to the right or to the left of 50%,
independently for either candidate. We ran the
simulator repeatedly, adjusting both positive and negative participation bias percentages for both
Bush and Kerry voters, until we were able to reproduce E/M's reported mean and median WPE.

That seems to say that they ran the simulator repeatedly, each time fine tuning the participation rates, until they achieved a reproduction of the 3 criteria across all partisanship bands. So it looks to me like participation rates are input to the simulator, not calculated by it.

Or, another way of saying it, they solved for the participation rates that they would have to plug into the simulator in order to get it to output results matching the actual data. This is basically an iterative solving process although one that is somewhat manual and based on eyeballing. So the participation rates are not the result of a simulation, they are the input to the simulation.

It seems that you and I interpret a key factor of the simulation in opposite ways. I think the response rates are input to the simulation and you think they are a result produced by it. Can you point me to some part of Bruce's explanation that leads to your conclusion? Alternatively, do you know whether Bruce or USCV has supplied the source code for the simulation? In the spreadsheet all I see are results.

[TruthIsAll]

373. My optimization model is formula and constraint driven to produce output.

Their simulation is a hunt and peck process.

You need exceptional patience and eyesight to read those scatter charts. And even if you do, there is no way of knowing which SPECIFIC precincts they are talking about.

This scatter plot is a total joke.

All we need is the raw data. In lieu of this data, I use
1) Bush 2-part vote 51.23%
2) weighted average response rate (53%)
3) weighted alpha (1.15)
4) reasonable input ranges for Kerry vote% across precinct categories
5) reasonable input ranges for response rates across categories
6) sufficiently wide ranges for precinct alpha

I derive the results and Kerry's exit poll percentages from this MINIMAL input. There is no manual tweaking involved whatsoever.

The algorithm does all the work. It comes up with non-biased results.
In a second.

Eomer, as I told them all, they can request any of the range constraints they would like. I would run the model and give them the results.

The bottom line is that one can understand my model. The calculations are obvious just by reading and checking the numbers.

The constraints mentioned above are input to the optimizer which comes up with a feasible solution instantly. No need to run thousands of trials.

Its very simple. To believe Bush won, you must believe that Bush voters were reluctant.

I believe the reluctant Republicans were Kerry voters. That means Kerry won by much more than the exit poll 51-48% result.

[OnTheOtherHand]

375. part I can answer, part I can't (yet)

I know from my own simulation experience (or common sense) that one doesn't necessarily have to input "Kerry participation rate = 70.4%" in order to obtain individual runs where it does. I suppose it is perfectly possible that he targeted a Kerry participation rate higher than 70.4%, although I doubt it.

Literally, we know that the 70.4% is a calculated output because we can see in the spreadsheet that it is figured as an average. But that begs the question what the input was. It's a fair question, and I will make sure you get an answer.

Because "participation rate" without sampling error tells us nothing directly about response bias, I am not too concerned about the specific input. Again, in the presence of sampling error, calculated participation rates can be very far from the actual response rates.

I think Bruce's simulation is most useful for informing some of our intuitions about variance. His sensitivity analysis addresses (but of course doesn't definitively answer) some questions that I hadn't even considered.

[LightningFlash]

392. You're wrong. One has to know the input....

To ascribe the output. In this case, "interviewers" had to go out there interviewing more democrats everywhere (not just in moderate bush, strong kerry, or strong bush districts) and the exact output variance reflects just that.

That is very likely not possible, nor what happened. While sampling bias did occur, it wasn't because interviewers went everywhere out of their way to find democrats. The theory dies even faster than the first RBR does, and it shows what length they obstruct.

The game is over--Fraud did occur, proven now in two states. All that remains to be answered is if it occurred in at least 10 or more of the other states, inflating the totals.

[Name removed]

275. Deleted message

Message removed by moderator. Click here to review the message board rules.

[tommcintyre]

288. I don't think you "practiced what you preach"

<You said:>
"...free from unwarranted speculation about personal motivations, from appeals to emotion and from personal attacks.

I would ask all my critics in this forum to hold themselves to the same standard - if they can."

Did you do that in THIS post?
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#374657

<My response:>
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x374482#374747

[sunshinekathy]

289. Dishonest statements are always destructive IMO.

Bruce said that:

QUOTE: Robust debate is never "destructive" in pursuit of the truth! A "volunteer scientific research project" is successful when differing opinions are expressed, not suppressed. ENDQUOTE

There is a huge difference between "robust debate" and dishonest misrepresentations of the facts.

Making false statements is always destructive in my opinion, and Bruce has yet to retract any of the false statements that he continues to make about our working paper, despite my repeatedly pointing out to him in emails that the mischaracterizations he has been making about our working paper were not supported by any sentence in our paper.

I have asked Bruce dozens of times to constructively find the exact sentence, or mathematical or logic error anywhere in our current working paper. He has never done so.

I informed Bruce repeatedly over a period of weeks that the mischaracterizations he made about our paper were not supported by any sentences in our paper that I knew of, and asked him to find the sentences in our paper that matched his characterization of it. He never did.

I have offered many many times via emails to work with Bruce if he would specify even one exact sentence in our paper that he did not agree with, yet he has never taken me up on my offer to listen to any constructive criticism he had of the paper.

I stayed up all night the night before the AAPOR conference trying to incorporate as many of the points of Bruce's last minute, paper that he insisted we substitute for the one all of us had been working on, trying to please him.

I can only guess what O'Dell's motivations are, but his actions have been and remain very destructive.

Distorting the truth is always destructive IMO and never has a good outcome and should be retracted.

People may want to know that our main two detractors, Mark Lindeman and Bruce O'Dell have not contributed any mathematics to any of our papers to my knowledge, although Bruce has contributed some programming work. Both Bruce and Mark seem to excel at writing and communicating.

Perhaps Mark and Bruce may just have a little difficulty understanding the math that Ron and I have done, and are emotionally attached to a position that is not supported by the facts without some pretty implausible assumptions about the size of exit poll variance that are not generally accepted.

There is a LOT of work to be done now at USCV and Bruce could be helping to push forward our work rather than misrepresenting our work and refusing to give us any constructive criticism of it.

Right now, we need to:

1. Finish the working paper that is truly groundbreaking.

2. Write recommendations for states on what election data to collect.

3. Write letters to state elected officials convincing them to request the data we need from their county election offices.

4. Write a grant proposal for designing the database to obtain the funding we need for the next phase of implementation.

Plus a gazillion other things. Bruce could be helping with those tasks and many others rather than trying to tear apart what we've accomplished thus far.

We've been trying ever since the AAPOR conference to get Bruce to behave constructively again, without success. Ron Baiman has been working very hard to show him the mathematics of it all, and pleaded over and over to Bruce to make constructive specific honest criticisms of our working paper if he didn't agree with something in it.

Best,

Kathy

[LightningFlash]

290. I was under the impression there were other motivations afoot....

Bruce seems to be toting the line of Mark Bluemanthal, and after the official "Census" results were repeatedly pointed out by persons, including myself, he has ignored that data substantially and not even uniformally addressed it.

As per my other observations, OTOH seems to do nothing but call the estimates a "faulty" sample. This has been proven wrong many times, that the U.S. government themselves conducted the sample and if true, they need to take it up with the U.S. government (Whom we know is not straight forward with us)

Therefore I assume there is an alterior motive of clear distraction, emanating from Mitofsky's supporters, and further confused and muddled by O'Dell's own motives as well as that of Mark Bluemanthal.

To this date they still refuse to answer their motives, nor answer the fact of the bogus mis-represented weightings. They either can not, or are simply distracting.

[autorank]

294. A fair prediction and a fair caution

The "free fall" in popularity for Bush represents a pathway to both salvation for our country and a resurrecting of the "fighting Democrats" we used to know. Bush is at 43% today, and that's with some artificial propping up. Maybe 5-15% of that number consists of people who just are not prepared to admit they voted for this monster. I don't blame them, they never intended to support someone this vile). As more of the speculated 5-15% wake up, and they will because they're as patriotic as the rest of us, Bush will drop to about 35%-38% popularity.

It is at that point that there will be demands for thorough investigations of all aspects of Bush corruption. One clear path to understanding how our nation was taken in will be a real look at election fraud. The compelling evidence brought forward on DU will look like a pre-cursor of those investigations. From voter suppression to rigged machines and tabulators; from stacked polls to attempts to "debunk" the obvious evidence offered by the National Exit Polls and other data archives, there will arise a story of fraud and deception. What more convenient and broad based explanation will the public need to understand how they were hood winked.

It is at that time that it will come time to pay the piper.

Whose side are you on?

What will the account of history be of your efforts based on your choices now?

How will you look your self and your fellow citizens in the face if you helped those who tried to "cover up?"

Fascinating isn't it.

"But many who are first will be last, and many who are last will be first."

Whose side are you on?

[Name removed]

305. Deleted message

Message removed by moderator. Click here to review the message board rules.

[autorank]

307. TIA must have hit the jackpot today,. You're positively apoplectic!

:rofl:

[foo_bar]

308. "First they laugh at you, then they fight you, then you win."

The truth rarely agrees with the Ministry of Truth.

[autorank]

309. A bit hostile tonight, are we?

You are indeed foo_bar

:rofl::rofl::rofl:

[foo_bar]

310. I'm only tellin' it like it is

"They just think it's hell."

[Melissa G]

312. foo_bar, What are you talking about on this thread? n/t

[Name removed]

313. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

316. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

324. Deleted message

Message removed by moderator. Click here to review the message board rules.

[foo_bar]

328. I agree about the cold handed theft

It was death by a thousand cuts. Touchscreens alone could have made the difference in Florida (if the anecdotes are extrapolated to the general population, IMO). The GEMS tabulators were clearly designed so that a computer-incompetent person could change the data without calling a technical person (i.e., Microsoft Access). Palast's report alone would change the outcome by ten times the margin of Ohio, so I have no doubt "we wuz robbed" by hook or crook, Ken or Jeb.

[LightningFlash]

330. Thanks for being reasonable.

Now, tell the right-wingers that and let's change this for the final time.

Obviously, the only thing there is to *disagree* about is the methodology used to MEASURE fraud.

So let me put it this way- I have no doubt the exit polls were *OFF* as per each state, by around 5-10%. This is not enough to make Kerry lose or anything else. It's just common, easy sampling bias. Nothing like some "incredibly faulty poll" that Mitofsky claimed to be as in his dissonance.

Okay? So what is the logical measuring stick...The electoral OPTIMIZER. It measures basis of fraud using TIA's expert calculations, and the maximum available level of "BIAS" for individual states.

It was derived from, in many eyes, the best stick to measure there is.

http://www.census.gov/population/www/socdemo/voting/cps2004.html

Lots may say, still, that government Census data is not enough but I think we if we are serious should probably trust the *official* Bureau of government statistics....Who basically include everything. This goes to prove just how many are dead voters, double votes, etc. etc.

It's not flawless, no, but the optimizer model of this gives us the best available picture overall of plausible and consistent machine fraud.

[TruthIsAll]

317. foo bar, this BINOMDIST is for you

How is the reconciliation project coming long?

Did you find any votes switched from Bush to Kerry other than the 2 that I came up with on EIRS? And what about Kerry votes switched to Bush? Any more than the 88 I found?

If you do, let me know. I will use your favorite BINOMDIST to recalculate the probabilities.

Meanwhile, we'll just have to shout it loud:

86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;
86 of 88: 1 in 79,000,000,000,000,000,000,000;

keep repeating:
first they laugh at you...

The great Mahatma must be rolling over in his grave knowing that YOU, of all people, quoted him...

BTW I'm shocked that you actually believe the election was stolen.
I guess that also means that you agree the pre-election polls and the exit polls that showed Kerry was a winner must have been right after all.

And my pre-election 99% probability calculation that Kerry would win was right after all.

And my 1 in 19 trillion calculation that 16 of 50 states would deviate to Bush from the exit polls was indeed correct, as well.

Congrats, foo-bar, for deciding to swallow your pride by joining the winning team.

[foo_bar]

319. can you provide a source for "86 of 88"?

i.e., not the EIRS home page. If you can't, that means the 79 quintillion is garbage-in, garbage-out (plus you rounded it to the nearest quintillion).

Did you find any votes switched from Bush to Kerry other than the 2 I found?

This one might be, or it might not be:
Caller claims that the machine had trouble staying on Republican candidates - it kept switching to Democrat like a magnet
link

Point is, saying "86 of 88" as though it were etched in stone is deceptive enough without ever providing a link to that effect.

(edited for broken link)

[LightningFlash]

320. Actually TimeForChange is the source. And you're wrong plain and simple.

[foo_bar]

322. source as in the place I can look at it

[Time for change]

323. Actually, I found 87 out of 94

vote switching incdidents reported to the national Electronic Incidence Reporting System favored Bush:

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=371211

[LightningFlash]

325. I saw the original report got removed. Do you have a new link up to it?

[Time for change]

326. Do you mean the Appendix and the Table?

If you do, you can pm me with your e-mail address, and I'll send them to you. I don't know what happened to the link. It worked before.

[LightningFlash]

327. Dissonance maybe? Things get hacked. In any case great report.

[Time for change]

331. Thank you. The link's on my post 326 above

It works for me.

[LightningFlash]

333. The link not working is on picturestable.com. N/T

[foo_bar]

329. what's the =BINOMDIST that we'd all be here Saturday night?

[Time for change]

334. If all 60,000 DUers were here it would be well over a trillion to one

But just three of us -- not very high.

Anyhow, I'm going to bed now, so that'll make it even lower.

[TruthIsAll]

336. No, I was the first. I found it months ago and downloaded EIRS

I applied Excel's data filter to extract the touchscreen vote-switch incidents.

This was the first time
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=126395

and I posted it again, because it is SO key.
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x369374#369410

[TruthIsAll]

332. Of course, here's a link (goto Post #7 - and add them up for yourself)

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x369374#369410

And this was the first time I posted it. http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=126395

[TruthIsAll]

345. I did so provide the link to EIRS. Read the two threads I started.

So now that you have had a chance to review the 88 incidents, what are your final totals?

Based on the FINAL numbers you will provide us, here is how we will compute the probability:

If YOU determine that N is the total number who switched, and that K switched from Kerry to Bush, we can calculate the probability that at least K of N would switch from Kerry to Bush as:

Prob (at least K of N votes would switch from Kerry to Bush) =
1 - BINOMDIST(K-1, N, .5,TRUE)

I am waiting for your final values of K and N.
Thank you for your honest efforts to calculate a final probability.

This one is very clean.

No MoE.
No cluster.
No bad polling.
No argument.

Just mathematics.
Unassailable mathematics.
Irrefutable mathematics.

[foo_bar]

337. responding to your edits from an hour later...

I guess that also means that you agree the pre-election polls and the exit polls that showed Kerry was a winner must have been right after all.

Concentrating so much effort on self-contradicting polls seems more like a quest for vindication than a quest for specific change; the problem is finding the hydra's nine heads and cutting them off, not "It's a real big hydra. 7 billion to one big!"

Not to mention that purged "felons" and otherwise disenfranchised voters didn't even make it to an exit poll (unless they were Exuberant Non-Voting Responders), so you're describing at best a subset of The Problem. In so doing, you employ a hypothesis that refers to the smoking gun and the source of a coverup as the same source; yet the election being BS and Mitofsky being BS aren't mutually exclusive. For example, Vermont had virtually no evidence of cheating of any kind, yet it's the second largest "red shift" (source, source)

And my pre-election 99% probability calculation that Kerry would win was right after all.

I'd continue to file it under GIGO, because you dropped polls that you felt were "biased" (in your word). That kind of adjustment is a good working definition of bias; whatever output you got was premised on a supernatural ability to detect biased polls.

(I personally thought the pre-election polls were underrepresenting Kerry, because no one in their right mind would tell a stranger on the phone they were voting for the opponent of color-coded terra alerts, an rKr hypothesis if you will.)

And my 1 in 19 trillion calculation that 16 of 50 states would deviate to Bush from the exit polls was indeed correct, as well.

Only if you assume a "random sample" (not a "random selection"), and only if you compensated for response bias as a spoiler of randomness (to assume Mitofsky's competence in fudging this correctly is to assume Mitofsky's competence). You can discount group psychology and language nuance at the peril of accuracy.

Congrats, foo-bar, for deciding to swallow your pride by joining the winning team.

I don't see it that way, but it's the closest thing to a compliment I'm likely to hear from you, so... cheers.
:toast:

(edited broken link)

[Name removed]

311. Deleted message

Message removed by moderator. Click here to review the message board rules.

[foo_bar]

314. where's Madame DeFarge when you need her?

[Melissa G]

315. Too much Testosterone! settle down guys...really. Come the

revolution the Women need to be in charge for a while...
Can we get back to the thread topic?

[tommcintyre]

318. Melissa, it was the "fish wives" who stormed Versailles ;)

and were very instrumental in securing France's democracy originally (With blades brandished, they decapitated the Swiss guards and put their heads on pikes). So, I'm not so quick to put the "women in charge" (just kidding ;) ).

And... WHAT exactly IS this thread's topic (agenda) anyway? Hmm....

[anaxarchos]

356. There you go again...

Steal a few votes, install a couple of judges and sure enough, you democrats are invoking images of the French Revolution... Sheesh...

BTW, this was not the only instance in the French Revolution. Someone (I think maybe Danton) summed it up as "The men talk; the women act."

I will enthusiastically vote for the women.

[Melissa G]

359. And I Hope your Vote will Be Counted! (grin) n/t

[foo_bar]

321. point taken.

[kster]

335. TIME FOR CHANGE IF YOU

"have very mixed emotions about starting this new thread" TELL THE MODS TO SHUT IT DOWN.

[autorank]

339. kster, It's dangerous to make too much sense to some people!

udaman!

[Time for change]

341. I don't think it should be up to me to shut it down

You might notice that I went several days without posting to this thread, hoping that it might go away, since it seemed to be upsetting some people so much, and since I was getting attacked for it so much -- which I really hadn't anticipated. After all, even the "opposing side" isn't even arguing for rBr, and mostly does believe that the election was stolen. The argument is simply about what evidence is solid, and therefore should be presented publicly.

So what I'm saying is is that I wouldn't mind seeing this thread go away, simply because it makes me personally feel very uncomfortable, given what I mentioned above, if for no other reason. But obviously, a lot of people want to continue to debate this, otherwise they wouldn't keep posting to it. So what right do I have to shut it down? And, would the administrators shut it down just because I asked?

And furthermore, I think that it's reasonable to believe that this debate has produced some positive effects.

[autorank]

349. Well, then :"just do it!" You CAN ask the mods to shut it down.

You say: "hoping that it might go away" and "I wouldn't mind seeing this thread go away,"

Well lets do it. The procedure is to simply go to the original message and click on alert. A little box will come up and you can make the request there in unambiguous terms.

Do you know that the "in the trenches" election fraud workers, the people who have put their names and lives on the line to fight the fraud, have actually started a thread called "FOR PETE'S SAKE STOP THE DEBATE BETWEEN O'DELL AND USCV!"
Click Here

Why would they want that? Because this is a distraction to their considerable work on the ground to fight election reform. This thread was NOT started by autorank or tommcintyre. Just a bunch of people who are irritated at the ad infinitum world with out end "debate."

You need to understand that there is a natural suspicion of motives when it comes to election reform organizations. People who might seem to be "friends" (to the position you staked out) may, in fact, have their own, separate agendas.

I looked at the site for the Election Sciences Institute (formerly votewatch) thinking it would be a good resource, particularly since it focuses on Ohio. Wow, what did I find: an article "Ohio Exit Polls ‘Not a Smoking Gun’ for Fraud, Study Says" presented at the American Association of Public Opinion Researchers (AAPPOR)where they argue against fraudin Ohio! Very confusing, not productive.

So here we have it, what's productive and what isn't?

This thread is well beyond it's productive value.

As Starskey says, Just do it!" Make it go away.

[Melissa G]

351. I asked twice and Mods did not shut the other Nice one down n/t

[kster]

338. Please get rid of

this b*llsh*t all you have to do is ask the mods.

[LightningFlash]

396. Allow me to clarify the point of county-to state level manipulation...

The exit polls while fitting, show there is a very unabashed discrepancy or disparity in the actual percentages for entire catergories such as Gender.

The result for this, is that these percentages "switch" places in an unaccounted for table. If column 1 was originally 51% with these responders ahead, and 48% with these responders behind, column 1a is now 48% with these responders behind, and the other responders 51% ahead.

This can not be directly explained by any so called "precinct to error" results. Purging in precincts is a common thing in a place such as, Florida, but can not explain that.

County level results are wired into a direct feed on the county level. If someone was playing on their computers, namely the BOE, then the vote columns will be switched.

All the votes that were accounted for will automatically be switched, much the same as Clint Curtis program. Clint Curtis program on the other hand, would make much clearer use in touch-screen central computers, explaining the over-encompassing "switching" votes.

Like in Blackwell's office, a county level re-write will overwrite all existing data by simply switching the columns. Not anymore than BOEs such as Tom Noe are needed for any of this, no payoffs, no hush hush....Quick switches, and the result is overturned.

Stolen election run like a wind up clock.

[Melissa G]

343. There is a nice Bruce O Dell thread over here

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x377355
If you wanna be Nice Please come visit...

[Time for change]

346. Very nice idea Melissa -- Thank you!

I recommend that anyone wanting to continue this discussion transfer it to the NICE thread.

[LightningFlash]

362. Point of order: Auditable machines AND recounts.

The bias idea has long since been refuted because of the proven manipulation in the voting machines.

Now, listen to my statement very clearly.

We either have openly auditable machines and recounts, or we have neocons electing themselves. All of the RINOs and DINOs will stay in power and turn this into a totaltarian state.

If you even for one minute slip up or doubt this, look at other countries now. The machines are being installed everywhere, in order to "protect democracy" but they are rigged by far-right extremists. They will PUT WHOEVER THEY WANT INTO POWER, IF THE MACHINES ARE NOT AUDITABLE.

That would end fair elections, that would mean no more republicans or democrats could elect who they want. The consent of the governed removed and instead--They elect themselves, NO THAT CAN NOT BE ALLOWED.

http://www.votersunite.org/

[Time for change]

366. I agree totally

That is not what this particular debate is about, although it is related. This particular debate is about how best to go about investigating the exit poll data. This is a very important issue, because thorough and scientifically proper analysis of this data is needed, in an effort to find, IF it is evident from analysis of the exit poll data, evidence of fraud. Bruce O'Dell felt that some members of USCV were going down the wrong road in that respect, and he was very concerned about that.

[Bill Bored]

374. Here's how:

I've been discussing this off the board but my idea is this; feel free to poke as many holes in it as you like:

1. Fraud would probably have occurred at the county level for various reasons. It's difficult (though not impossible) to hack specific precincts, but much easier to swing entire counties.

2. Most counties have both Bush and Kerry precincts, but in Kerry counties, Kerry precincts would predominate. There would still be some collateral Bush precincts that would have his margins enhanced, but these would be in the minority.

3. According to the slope debate and the distribution of WPE by partisanship, it seems as if fraud would have been carried out in Kerry counties. Bush counties could have been left alone (if it ain't broke, don't fix it!). Also, there aren't enough strong Bush precincts to change the outcome of the election anyway.

4. Strong Kerry precincts were mostly in urban areas, which could comprise entire counties. While there are a fair number of these precincts in the exit polls, it's possible they were mostly left alone too because administration in the cities is a bit more competent and bi-partisan than in the country and the suburbs where there's more outsourcing of the electoral process. So you end up with over twice as many strong Kerry precincts as strong Bushes.

5. Could county-level vote switching (or vote deletion) be modeled or inferred from the exit poll data which has only precincts, but not counties? It could explain why we see so many middle and moderate Bush precincts, but fewer moderate Kerry precincts, and even fewer strong Bush precincts.

6. Personally, I think any models that don't take an educated guess as to the methodology of the fraud are a waste of time. What do the rest of you debaters think? (And if there's general agreement, can we start a new thread on this subject? This one's getting kind of long!)

[Time for change]

379. I think that a new thread on this issue is a very good idea

But as long as I'm here now, I'll take a wild shot at it.

I agree that it would be far better if we had a model that takes an educated guess at methodology. But unfortunately, I've heard a lot of contradicting theories, but nothing that could be used to test the exit poll data. The reason for this may be that the cheating was done in many different ways, or simply that nobody has yet discovered how it was done.

I heard this idea at a conference that I just attended: That as long as we could get a final count at a given precinct when the precinct closed, it would be much more difficult for that precinct to be affected by fraud. That would be because the Republicans would be much less likely to change a count if they knew that someone already had the real count. But I don't know how that could be used to test exit poll data.

As far as the other stuff is concerned, I really don't think we know at this point whether most of the fraud occurred in the Kerry, Bush, or bipartisan precincts. I think there are reasons to believe the two former possibilities, so the evidence is conflicting as far as I can make out.

[Bill Bored]

383. I disagree

You wrote,

"as long as we could get a final count at a given precinct when the precinct closed, it would be much more difficult for that precinct to be affected by fraud."

I wouldn't say much more difficult. And this goes to the heart of the analysis. WPE is error WITHIN the precinct. Whether it's a singleton (one precinct), a sample (a bunch of precincts) or a statistic (a mean or median derived from the sample), the errors (whether poll or vote count) still occurred within the precincts themselves and would therefore not be detectable by verifying precinct totals alone because the totals would already have been corrupted. To change them at a central tabulator, as seen on TV, is a crude methodology, and as you say, it is detectable.

If we're saying there is within-precinct error, based on the exit polls, we need to look at how this could have occurred. And there are ways besides the exit polls being off.

Comparing precinct totals to centrally tabulated totals is a good way to detect hacking of tabulators (a la Bev Harris' TV demo) but it would NOT detect true WPE.

I believe there are ways to exploit e-voting software to enable WPE to be distributed throughout a jurisdiction (county) on lots of machines as the votes are cast.

As far as which counties, one way to arrive at the current distribution is to shift votes toward Bush in the Kerry and the bi-partisan precincts. As Febble says, up and to the right on the scatter plot. (Reminds me of Jim Garrison's JFK assassination case: "Back and to the left." but this is "Up and to the right.")

[Time for change]

384. Look at it like this

A certain precinct votes 50% Bush, 50% Kerry. The polls also show 50% Bush, 50% Kerry. If no fraud, WPE = 0.

But no count is recorded, and when the count comes out of the central tabulator, the vote is 70% Bush, 30% Kerry. The WPE for that precinct would be -40.

So I don't understand why you're saying that the WPE wouldn't be affected. Or maybe I misunderstand what you're saying.

[Melissa G]

385. County manipulation would Blur WPE

edit to say. If I have it right BB would be saying all this back painting of WPE is irrelevant because the constuction of the fraud is off a different theory or model.

[Bill Bored]

387. Not exactly.

What I'm saying is that at the level of the county election system, there are ways to influence individual precincts en masse. So you could measure errors at a bunch of individual precincts but they could be generated from a central point either before of after the election.

If after, they would probably be just tabulator errors and I would not characterize them as being within precinct. This is the Bev Harris hack.

But if the count is altered before the election (by configuring the machines to count incorrectly), it would truly appear as errors made within each individual precinct.

TFC has suggested that the term WPE does not even necessarily mean the error was made within the precinct, but rather could be a simple tabulator error.

I tend to think it means what it says, i.e., that there is a discrepancy between the actual precinct percentages and the exit poll.

[eomer]

391. Definition of WPE

Hey Bill, WPE can't possibly mean what you are thinking it means. For it to mean that would require that E-M had access to the official count of each precinct somehow at the precinct level, before the votes are processed by a county tabulator. This information is simply not available in many, probably most, cases.

I believe the term WPE as calculated in the E-M report has to include all errors in the official counts that occur anywhere in the process up to and including the point where E-M receives them. And it has to include all errors in the exit poll samples that occur anywhere in the process up to and including the point where E-M receives them from the poll taker. This simply has to be the case because they would have no way of calculating WPE by any other definition.

WPE even has to include any typos on the part of some E-M employee when transferring a precinct-level official count or exit poll sample into their system.

The question then is not one of definition. It is a question of knowing how E-M collected the precinct-level official count numbers. I don't believe they would have gotten the official count numbers directly from the polling place, even in those cases where it was possible. I believe they would have gotten these numbers from the county or from the state.

The question of whether WPE would include tabulator fraud would then depend on exactly where E-M got their information from and also on exactly what kind of tabulator fraud we are talking about. If a county tabulator modified each precinct-level count before they were rolled up and it reported out those fraudulent precinct-level counts then WPE includes that error. If the tabulator reported out true precinct numbers but then fraudulently added them up into county or statewide totals then WPE does not include that error.

The question of whether or not WPE, as calculated by E-M, reflects tabulator fraud would therefore have to be answered by looking at the specific details within each jurisdiction. The answer may well be different from one jurisdiction to another.

[Bill Bored]

402. Right.

And this is why the whole exercise may be futile.

However, could it explain why the WPE went both ways? (I haven't had time to sort that out yet but it's obviously an important consideration which has been ignored by most of the true believers among us.)

Any theory of fraud must take into account the large WPEs in both directions. I know one way this can be explained, but it relies on county level manipulation of precincts at the precincts. Is there another way? Maybe.

[LightningFlash]

393. Bill Bored has it correct. The fraud was done county level.

Purging was done in the precincts, but actual "manipulation" is conducted at county level.

Confirmed by Bob Fitrakis, Lawyers in OH
http://www.bradblog.com/bradshow/

It wasn't that much of a tidy effort. Two way re-writable software was installed, which manipulates the count at the county level. Blackwell had it wired into and controlled from his office.

Precinct level is impossible, there can only be "various problems" of correcting procedures....

County level, vote manipulation is a reality.

[Bill Bored]

401. Read my posts again.

I said something like:

The precincts could have been manipulated en masse.

This means that at the county level (administratively) you can distribute the error to the precincts. (See Appendix 1 of Bruce O'Dell's paper for a summary of the problem.)

In the case of Blackbird, you are really talking about the State level distributing fraud to the counties. This may have been the best they could do in Ohio due to limitations of the punch card technology. But with e-voting, it can be done one layer below.

In Ohio, you do have the ballot order rotations in which Bush and Kerry punch cards were lined up to allow vote switching between them. This would be within-precinct error. There is NO good reason to line them up that way. If they lined up Bush and Kerry only with third party candidates, fraud would be detected immediately, but NO...they had to align Bush with Kerry in collocated precincts. I hope some of the many investigators in the state will come out with a critique of this that will point out the potential for manipulation, even if unproven.

[Bill Bored]

386. Because the exit pollster looks at the precinct total and compares it

to the poll result to obtain the WPE. The tabulator result is aggregated with all the other precincts in the county and presumably is never even seen by the exit pollsters on election day.

Otherwise, why even use the term "Within Precinct Error?" If it's even possible that what the WPE is measuring is a tabulator error, i.e., an error made BEYOND the precinct, then it would be a misnomer to call it WPE. I think it should be given another name or perhaps I'm taking it too literally. I had always assumed that the term "within precinct" meant just that. If not, then I'm REALLY getting sick of these exit polls!

In any case, if the error (poll or vote count) really is made within the precinct, it would certainly show up as "WPE."

It would be important to characterize this better though.

[eomer]

394. Bill, see my post #391

I don't think it is possible that the exit pollster calculates the WPE. In many cases there are no results whatsoever available to the public at the polling place. Another reason it can't work that way is that many polling places would still be open at the time of the final report from the pollster to E-M. That cannot possibly be how it works.

So WPE cannot mean only errors that occurred within the precinct. That would be impossible to know.

What I believe E-M means by WPE is an error contained within their precinct numbers (however obtained), before they apply the weighting.

[Febble]

395. See my PM n/t

[Bill Bored]

400. If true then perhaps we are all wasting our time.

If WPE can't even be defined consistently, how can it be analyzed?

Tabulator errors could be found and corrected or avoided in jurisdictions that do precinct canvassing, and not in those that don't. So now you have yet another variable to consider!

This Mitofsky guy is really starting to look like a tyro when it comes to voting technology. (I know this criticism has been raised by USCV re the Jan. report which did NOT distinguish between pre- and post- tabulator "WPE.")

This is perhaps yet another reason for looking at county-level discrepancies.

[Time for change]

404. Here's a shot in the dark

Regardless of who's right and who's wrong about this WPE discussion,
IF Mitofsky has final precinct level data that came out of the precinct, and NOT the central tabulator (which I have to admit never even crossed my mind before this discussion), could that be compared with the central tabulator numbers??? If there are big and frequent discrepancies there, that would prove fraud.

[eomer]

408. I agree that would be very useful information to have.

It never crossed my mind either (amazing that this long after the fact we can come across something this basic that many or maybe all of us have not considered). If this polling place count information does exist it may contain a smoking gun when compared with the final official count.

I should say I think BB's definition may be right in the majority of precincts. There are some precincts where a polling place count exists and others where it does not. I don't know what the percentages are.

Any precinct that uses opscan with centralized scanning would not normally be counted at the polling place. I don't know how many opscan precincts use centralized scanning and how many use polling place scanning. Also, I was under the impression that most DRE systems did not have a count at the polling place - that the memory cards were transported to a centralized county operation and were not counted until arriving there. Does anyone know how many DRE systems include a count at the polling place?

Another question is whether a polling place count that exists is then made available. I believe that some jurisdictions require it to be posted on a door. But there may be places where a polling place count is done but it is secret. Are there cases where a count is done but it is secret unless you are someone special (E-M for example)?

[Time for change]

417. Please take a look at this thread

I'm trying to transfer this part of the discussion to another thread:
http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x378357#378360

I'm very intrigued by what you have to say about this, but I'd like to transfer the discussion to another thread, because this one is too large, too many different topics are being discussed, and also some people are very upset about the fact that this thread is persisting, given the fact that it emphasizes dissension within USCV.

Could you please go to the above thread and see my post # 18. Thanks.

[Bill Bored]

420. What's the record for posts in a single thread anyway?

I thought we had a shot at it but I guess it's time to move on.

[Melissa G]

426. This is still a long way from a KOEB thread in the heyday!

500 plus posts...No bodice ripping of EDV's (that's Estrogen Driven Vixens)here...

[Bill Bored]

418. Not amazing -- to be expected.

eomer wrote:

"It never crossed my mind either (amazing that this long after the fact we can come across something this basic that many or maybe all of us have not considered)."

I've been considering it for months. The reason why it hasn't come up is because all this exit poll stuff never looks at HOW the fraud could have been done. IMO, THAT'S where to start -- NOT with the exit polls.

In fairness though, this issue was raised in one of USCV's earlier papers, or perhaps it was Freeman's.

What's he been up to lately BTW? Has he taken a side in the BIG SLOPE DEBATE? Has he gotten a haircut yet? (Just kidding. I'm a kidder you know.) :)

[kster]

390. My Condolences to the family no# 1708

http://www.centcom.mil/CENTCOMNews/Casualty_Report.asp?CasualtyReport=20050618.txt

[kster]

388. Heres A Reason To Keep Fighting The Freepers !

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=104x3858119

[LightningFlash]

397. Another quick shot analysis within here.

http://www.exit-poll.net/election-night/EvaluationJan192005.pdf

The following table is able to show us clear manipulation and problems in the state of florida. So we have a state that uses memory cards, "stolen" votes, purging, and finally switched unver-votes.

If one could derive a table from the following model, that could correspond to each state, we could discover which type of manipulation and so on is going on in each state. It would measure it accurately in the case we know machine type, day of issue, race, age, and other statistics especially in the larger more "obvious" places like Wisconsin.

It's a great place to start building a map.