Next time, you TIA-bashers, bring YOUR own model into my thread, OK?

[TruthIsAll]

Hey, bring it on, all you guys who love to bash my probability posts without supplying a model of your own. You know who you are. Come forward. Put up or shut up. I can take it. Tear my logic apart, but do it constructively. Don't make an ass of yourself by offering nothing but bile. You do all of us a disservice by your mindless diversion. Almost makes me think that I am getting too close to the TRUTH and in so doing makes some here uncomfortable.

In fact, those of you who are the first to criticize never seem to start new threads with an original idea; you only react by coming out of the woodwork as if on command. You guys are awfully fast to respond to a certain type of analytical post. Kind of makes me want to reach for the tinfoil.

Jeez, this is supposed to be a Forum for people to work together in shedding some light on the Bush Cabal, not for tearing each other apart.

Say what you will about my attempts to shed some circumstantial light on the 2002 Election fraud or the JFK assassination, at least I DO THE FUCKING WORK. And so far, these posts have been for the most part positively received by people like Bev Harris and ProfessorGAC.

I never do an analytical post pretending to know all the answers. They are just a start; I will always welcome alternative logic or criticism.

It's funny, those who respond with their own analysis usually agree with mine. Those who just use sarcastic one-liners in attempting to belittle my work never accept the challenge of providing their own analysis. You know who you are.

It's time to put up or shut up.

[Terwilliger]

1. TIA, you do a shitload of work

I'm proud of ya :bounce:

[JustJoe]

2. What posts on the JFK assassination?

I've only seen passing references to it
since I've been here.

[TruthIsAll]

4. Here it is...

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=104&topic_id=6304

[TruthIsAll]

24. Here are the JFK mystery deaths referenced above. The link is broken .

11/63 Karyn Kupcinet TV host's daughter who was overheard telling of JFK's death prior to 11/22/63. Murdered.

12/63 Jack Zangretti Expressed foreknowledge of Ruby shooting Oswald Gunshot Victim.

2/64 Eddy Benavides Lookalike brother to Tippit shooting witness, Domingo Benavides Gunshot to head.

2/64 Betty MacDonald* Former Ruby employee who alibied Warren Reynolds shooting suspect. Suicide by hanging in Dallas Jail.

3/64 Bill Chesher Thought to have information linking Oswald and Ruby Heart attack.

3/64 Hank Killam* Husband of Ruby employee, knew Oswald acquaintance Throat cut.

4/64 Bill Hunter* Reporter who was in Ruby's apartment on 11/24/63 Accidental shooting by policeman Death judged non-suspicious by HSCA.

5/64 Gary Underhill* CIA agent who claimed Agency was involved Gunshot in head ruled suicide.

5/64 Hugh Ward* Private investigator working with Guy Banister and David Ferrie Plane crash in Mexico.

5/64 DeLesseps Morrison* New Orleans Passenger in Ward's plane Was mayor of New Orleans.

8/64 Teresa Norton* Ruby employee. Fatally shot.

6/64 Guy Banister* x-FBI agent in New Orleans connected to Ferrie, CIA, Carlos Marcello & Oswald. Heart attack

9/64 Jim Koethe* Reporter who was in Ruby's apartment on 11/24/63 Blow to neck.

9/64 C.D. Jackson "Life" magazine senior Vice President who bought Zapruder Film and locked it away.

10/64 Mary Pinchot JFK "special" friend whose diary was taken by CIA chief James Angleton after her death. Murdered.

[oasis]

3. I appreciate the work you do, and that of Bev and DemActivist also.

The continued success of DU will depend on contributors who are willing to roll up their sleeves and do the research.

[Old and In the Way]

5. No criticism from me.

Hell, I can hardly understand the math.....

[Cheswick2.0]

6. TIA, I barely got a B- in a 100 level stats class

I salute you for all that you do here.

[treepig]

7. i think the basic problem is a fundamental misapplication of statistics

the models you present are fine for amusement purposes, but it's a huge stretch for them to have any further significance (such as establishment of culpability in any meaningful way)

the main reason is that there are an incrediblely large number of events going on, and statistically improbable happenings take place on a regular basis. for example, what are the odds that i would eat the exact tomato i ate for dinner yesterday? i'd estimate about one out of 10 billion, yet it happened.

or what are the odds that bill clinton became president? there were at least 100,000,000 people eligible for the job (u.s. citizens older than 35) so the probability that he would become president was .00000001 - MASSIVE fraud must have been involved for such an unlikely event to occur!

[TruthIsAll]

9. Its amazing how those truly statistically challenged, like yourself...

offer up criticism which belies your knowledge, or shall we say, lack of it. You are what I call a coincidence theorist, which just proves your lack of understanding of what probability theory is all about.

You say:
"i think the basic problem is a fundamental misapplication of statistics"
the models you present are fine for amusement purposes, but it's a huge stretch for them to have any further significance (such as establishment of culpability in any meaningful way)

I say:
You are partially correct. I DO enjoy the challenge. But, please, don't dare go to Atlantic City expecting to win. Any slot-machine playing grandma knows more about the odds than you do. I plead guilty of analyzing what you would call a "coincidence" in a way that would be obvious to anyone who has taken Probability 101 and applying it to the real world. The results are NOT for amusement, however. They serve to establish a circumstantial case for conspiracy using mathematics to explain what you would just call a "coincidence". To say it is a coincidence is to take the easy way out from your having to actually ANALYZE the problem, which would be beyond your level of mathematical maturity.

You say:
the main reason is that there are an incrediblely large number of events going on, and statistically improbable happenings take place on a regular basis. for example, what are the odds that i would eat the exact tomato i ate for dinner yesterday? i'd estimate about one out of 10 billion, yet it happened.

I say:
This is a very sad statement. Shows a complete innocence in the basic probability concepts. Let me explain and hopefully remove the web of confusion you are suffocating in.

Lets say I gathered all the tomatoes in the word and numbered them from 1 to 10 billion. I then put them into a large sack. And I told you I would give you a $10 billion if you picked out the one with the number 6,666,666,666 by reaching your very long arm into the sack without looking. The chances that you would get THAT one is 1 out of a 10 billion. The chances that you would get ANY tomato is 100%.

Are you any wiser now? Or just feel embarrassed that your ignorance of basic probability has been exposed for all of DU to see?

You say:
or what are the odds that bill clinton became president? there were at least 100,000,000 people eligible for the job (u.s. citizens older than 35) so the probability that he would become president was .00000001 - MASSIVE fraud must have been involved for such an unlikely event to occur!

I say:
This is exactly the same as above. Instead of a tomato, substitute Bill Clinton.

I am burning a few calories responding to your post. This is surreal. Do yourself a favor. In the future, to avoid looking...um.. silly, make sure you know you have a modicum of knowledge on a subject before commenting on it.

As Al Pacino playing Ricky Roma in "Glen Garry, Glen Ross" said speaking to his manager, played by Kevin Spacey, who has just fucked up Pacinos real estate sale:

"The next time, you child, unless you know what you're talking about, just keep your f****ing mouth shut!"

[treepig]

11. hey, have fun!

proving that statistically unlikely events take place is easy enough.

like the 18,181 - 18,181 - 18,181 vote total, that one's outrageously fishy at face value, if you wish to "prove" it statistically, well OK!

but it has absolutely no meaning without establishing cause and effect. without establishing cause and effect, you can't get things published in scientific journals (an area i'm familiar with) - your thesis would immediately be rejected as "speculation." i suspect a similar result would occur in a court of law (although i'm familiar with anecdotal evidence of strange goings-on in that venue, so who knows).

in any event, if you felt the need to debunk the outrageous examples i gave, that would seem to show an underlying insecurity on your part about the validity of your methods. just to be clear, i find your statistical correlations to be highly entertaining and encourage you to continue - but i'm just a bit confused as to your motive - are you here to entertain, or do you actually take yourself seriously?

[TruthIsAll]

12. I have just refuted YOUR twisted logic, but like a Repuke,

you ignore the TRUTH of what I have said and just attempt a further diversion.

I don't have to publish SHIT in any scientific journal if I want to present something here at DU.

Go pick a tomato.

[treepig]

14. ok, thank you for the clarification!

n/t

[TruthIsAll]

15. My motive? You want a motive? Um.... ok

"in any event, if you felt the need to debunk the outrageous examples i gave, that would seem to show an underlying insecurity on your part about the validity of your methods. just to be clear, i find your statistical correlations to be highly entertaining and encourage you to continue - but i'm just a bit confused as to your motive - are you here to entertain, or do you actually take yourself seriously?"


What a condescending statement! Do I take myself seriously? Yes. What is my motive? To enlighten, educate and entertain. Just like I am by those wonderful posters at DU who sit down, get to work and produce.

My only feelings of insecurity are for my country when I see those millions like yourself who have not an iota of scientific inquisitiveness, or sense of history, yet are always at the quick to ridicule those who do.

I draw upon many years of analytical, model building and programming experience; not to mention 2 masters degrees in applied mathematics which provided me with a reverence for the power of mathematical analysis in solving all kinds of problems.

I bet you have never taken a single course in Probability. Correct me if I am wrong.

[treepig]

18. i've taken statistic classes

and i'm not disputing your math, just its application.

consider an election where a candidate gets the following results:

precinct #1: 18,181 votes
precinct #2: 18,181 votes
precinct #3: 18,181 votes

total votes = 54,543 (to defeat an opponent by a narrow margin)

immediately, red flags are raised, and impeccable statistical analysis shows there is essentially no way this result could have been achieved by chance.

now consider the same election with the following results:

precinct #1: 21,477 votes
precinct #2: 12,182 votes
precinct #3: 20,884 votes

total votes = 54,543 (to defeat an opponent by a narrow margin)


is there anything suspect about the second election that would trigger a detailed statistical analysis to prove fraud took place?


ok, time's up - most people would say, no probably not! but me, being concerned for the good of the nation, looks further and finds that the person just elected has three children, with the following birthdates:

child #1: February 14, 1977 (2-14-77)
child #2: January 21, 1982 (1-21-82)
child #3: February 8, 1984 (2-08-84)

now what are the chances that this candidate, by chance, would receive the exact vote totals that correspond to the birth dates of his or her children? i could do rock-solid statistical analysis to show that there's no way in hell this could have happened by chance. very entertaining indeed, but in the absence of more evidence (perhaps as someone else suggested, proving this odd statistical event was delibrately programmed into the results as a 'clue'), then at the end of the day, all we have is entertainment.

[TruthIsAll]

19. You miss the point. We are not trying to "prove" anything..

We are just trying to deteermine the probability that the events occurred purely by chance - and then investigate further. This is a distinction often missed.

OK, you've taken Statistics courses, but what about Probability 101?
Are you familiar with the Poisson Distribution and its relationship to the well-known Normal Distribution for computing the probability of rare events?

I don't know if you have seen the analysis I did of the 2002 senate elections. Four out of 10 critical elections ALL fell for the Repukes, and in each case the final polling margin of error was exceeded by significant amounts.

I calculated the odds of this being due to chance at less than 1 out of 43,000. This was BEFORE the Bev Harris revealed the last-minute Diebold-Georgia software "patching" of 22,000 touch screen computers on the weekend before the election. Did I prove anything? No, I did not. Did I establish a mathematically circumstantial case for investigation of fraud, which was already felt to be the case by many? Yes. Even the well-known pollster Zogby said he had never seen the results of last minute polling so skewed.

By your logic, we should not even bother to analyze these strange events from a mathematical standpoint. Well, trial lawyers would disagree, as well as medical researchers.

Lets look at your example. Suppose a reporter became aware of the significance of those birthday voting totals. Suspicions would be raised, I am sure, if it turned out that there was also a relationship between the voting machine manufacturer and the candidate.

The reporter would, if he knew the odds of the "coincidence", investigate further to see if there was in fact such a relationship, based on these three apparent "coincidences".

So you see, the analysis has merit in any case, even if just to raise the suspicion of fraud.

Just look at what happened in the 2002 Georgia senate race.

[Cocoa]

21. the statistical analysis is not necessary

the anomaly is obvious, the stories about this quote the election workers as immediately seeing this result as freaky.

Part of my work is data quality, and I have a statistical background, and I have never once sat down and calculated a probability like this.

By far the most useful analyses are empirical, and in this case there's tons of empirical data available, tons of current and past election data from all over the country.

Instead of calculating the probability of this happening, why not go see how often this kind of result has actually happened?

[treepig]

27. you seem to be preaching to the choir here

i think most readers of this forun don't need detailed statistical analysis to be convinced something is amiss. they're already suspicious of fraud. but if, as you say, you wish to post stuff for du'ers to read, fine, i read it, and as i pointed out it's quite entertaining. but i'm already convinced of your premise, so big deal.

you also mention you're concerned about the country and want to do something about the impending loss of democracy, in this case you need to convince people who are not 'the choir' - and the standard is much higher. and if you spend time on doing all this analysis, don't you aspire to a higher standard so all your hard work, and stellar qualification in statistics, are not wasted on somebody like me who probably doesn't even know what a poisson distribution is?

you mention medical research so i'm sure you're familiar with dna microarrays, a technology that can give you a read-out of the expression of each of the 30,000-odd genes in a human cell. five years ago this field seemed to be much like the after-the-fact analysis of election results - there was vast amount of data generated. while this data was vastly entertaining for the people doing it (wow - look at the different sets of genes expressed in skin cells compared to heart cells!) scientific funding agencies such as the national institutes of health basically wouldn't sponsor this work, because after-the-fact looking for statistical correlations wasn't considered to be hypothesis-driven research (i.e., it wasn't 'science'). proponents of dna microarray technology were treated much more rudely than you have been, and guess what, they're better for it.

that's because the dna microarray advocates didn't get all huffy and accuse their detractors of not understanding their technology. instead they learned how to use the statistical correlations their algorithms generated for useful, predictive purposes. for example, let's say there's 100 cases of breast cancer and to a physician the disease appears to be identical in each case (and the physician would be likely to assign the same treatment strategy to each woman, a few would respond well, most not so well). however, upon analysis of gene expression by dna microarrays and statistical correlation of the results, it turns out that these 100 cases of breast cancer are actually 20 distinct diseases. furthermore, the dna microarray data can be used to predict which treatment strategy will be most effective for each of these distinct forms of breast cancer.

[TruthIsAll]

28. You'd be surprised who is attending here: many non-believers

We seem to be coming around a little bit, I sense. Good. Perhaps I am preaching to the choir NOW, but last January you would be amazed how many of those who are now "converted" believed the Repuke propaganda that the Dems lost the Senate because of "poor" turnout. I like to think that my analysis got more than a few thinking that perhaps there was something more nefarious going on.

As far as using the results for predictive purposes, I think it is fair to say that we are doing just that. We are trying to stop the GOP, Bush and the DIEBOLDers from running over our democracy in 2004 - like they did in 2002.

I can make THIS prediction with 100% CERTAINTY:

BUSH AND THE REPUKES ARE GEARING UP FOR THE MOST MASSIVE ELECTION FRAUD IN HISTORY. THEY KNOW THAT THERE IS NO WAY IN HELL THAT THEY CAN EVER WIN A FAIR ELECTION.

PAST IS PROGUE. WE ARE DOOMED TO HAVE ANOTHER ELECTION FRAUD UNLESS WE DO SOMETHING ABOUT IT NOW.

HERE AT DU, BEV HARRIS, DEMACTIVIST AND ELORIEL, TO NAME JUST A FEW, ARE DOING JUST THAT.

I LIKE TO THINK I HAVE ALSO CONTRIBUTED IN MANY POSTS, NOT JUST THE PROBABILITY POSTS, BY RAISING AWARENESS OF THE ONGOING CONSPIRACY TO STEAL OUR DEMOCRACY.

[treepig]

42. i didn't have time to finish the post yesterday

and left the analogy dangling.

what i intended to point out was that the medical researchers were successful because they used all the data for their after-the-fact analysis. the same approach must be applied to elections - criteria must be established and then strictly adhered to. it must be used to blindly analyze the thousands of state and federal elections that have occurred over the past 20, 30, or 50 years. i would accept that this type of analysis will identify 'suspicious' races worthy of further investigation. i will not belabor this point since others have made it much more eloquently than me by this time. nevertheless, from what i've seen here, the analysis has been completely backwards - identify races that were suspicsious based on a hunch (i.e., the odd 18181-18181-18181 connection or the fact they were deemed crucial for the republican takeover of the senate); only then were statistics used to confirm the suspicions.

and i apologize if i've been unduly ornery - it's probably due to the fact that i work in a lab, and i've found it work best if the toughest critics are fellow members of the lab, that way by the time it's time to present results to the outside world, it's a cakewalk - we can instantly and easily answer any and all critical comments (but it admittedly can be a bit rough on the thin-skinned)

[Nazgul35]

31. Actually...the poisson distributions...

is used when you assume independence and consistent intervals between the occurence of each event. From what I was able to pick up here, you are in fact assuming that events are conected to each other and that they do not occur uniformly across time...if this is true, then the sampling distribution you are looking for is the negative binomial distribution....

It seems, and please correct me if I ma wrong, like you are condoning data mining...which is a no-no in social research...your theory comes first, then you construct a model based upon assumptions and theories about the real world (and dependning on how specific you want to be when you generalize...the model resembles the real world closely...or vaguely) and then you use the appropriately selected sampling distribution (it connects the population and the sample based upon the theory you have outlined AHEAD OF TIME)....and test...

There are several levels of tests...with the lowest being what we call the classic linear regression model (where we use the Normal distribution for everything...completely wrong in most cases)...the next level deals with Maximum likelihood, where you use probability distributions to explain the likelihood that the data has the statistic (mean, meadian mode, variance, etc.) you have expressed prior to the test...the highest level is Bayesian statistics, which assumes the data is fixed and that the parameters vary....you set a prior value (based upon theoretical concerns) employ maximum likelihood using the correct sampling distribution, and see what you posterior distribution looks like....we use Markov Chain Montee Carlo techniques to describe the distribution we have found in the sample to run our tests...

Above all...no crunching of number means anything if it is not done with a prior theoretical hypothesis that builds a proper model and the appropriate selection of the sampling distruibution....

otherwise you are just data mining...patterns can be found in almost anything...consider the distance that the planets are from the sun....the distance of each planet is a certain factor (can't remember the exact number...maybe someone could remember...from the sun...) but that doesnt explain WHY the planets are those distances...and to date, we still have several theories about why, but none have been proven...

[TruthIsAll]

37. Nazgul35, yours is an incisive, intelligent post..

Yes, I am assuming independence of events, which is obviously not 100% the case. And your arguments for using those other statistical models/tests make a lot of sense. It is obvious that you are quite knowledgeable in statistics and probability analysis.

But I ask. Are we overcomplicating the basic problem?

Please comment on these specific assumptions:

We have N elections in a given county.
In 3 of the N elections, the Republican vote is exactly the same.
What is the probability of this occurrence?

Assume from past data that the expected Republican votes are distributed around an expected mean of 18000 and have a historical range from 17,500 to 18,500.

Lets assume these results are normally distributed as a first approximation. In fact, I do not have access to the historical data, so I merely attempted to give a quick approximation, based on these assumptions.

To my mind the problem is almost, but not quite, analagous to buying a lottery ticket every week numbered from 000 to 999. Here we should assume the numbers are uniformly distributed - not a Binomial (or Poisson) distribution.

Assume we play the lottery for N weeks, and the winning number, say 666, occurs exactly 3 times. What is the probability?

In our problem, the range and statistical distribution of the Republican votes for this county in prior elections are as yet unknown. I did not attempt to go full bore into the full analysis. You are quite correct that the underlying assumptions and distributions are not fully understood. But we can make a quick first-cut approximation.

I appreciate the obvious expertise you bring to the discussion.

[dsc]

38. The correct question is

what is the probability of any one number appearing three times. Presumedly you would have found any margin that appeared three times to have been suspicious.

[TruthIsAll]

39. dsc, precisely...

I used 666 as an example. But the true question is what is the probability that ANY result occurs exactly 3 (or more) times in the set of N lotteries or elections? Here it is more complicated, because we do not know precisely the number of Repub voters, the underlying distribution around the mean, the range, etc.

I think that perhaps the uniform distribution, where ALL vote counts are equally likely within a range, may be a good, first-cut model, due to its simplicity. In fact, this is what I used intially, until I meandered into the Poisson Distribution - which may or may not be appropriate.

[dsc]

40. I would probably start

by looking at the average percentages for Republican and Democratic candidates in elections going back some amount of time. That would give you a range of possible margins. Then you have to decide what to use over that range.

[Nazgul35]

41. to naysay the naysayers...

this actually offers a natural experiment...Our assumption is that Diebold, or other companies are rigging the election machines to ensure that certain results happen...one could test the hypthesis by examining the typical election results prior to and after the introduction of these type of machines...if prior to the intro, the occurence was not statistically significant than what nature would do, but after, we had a statistaclly significant result...you would have some coorboration on your premiss...

Of course you would (and by you, I mean anyone who would like to take up this challange) have to control for other possible causes and look for correlation...but you would have something there...

One Note: In probability, the prediction for a single point estimation is always zero percent probability...instead, one should examine the total number of election results that we would expect...if nature selected, versus what would occur if something (or someone) was screwing around with stuff...

I really don't think the arguments here are that far off...it's all a matter of methodology, some people believe that research should be driven by theory...not data patterns....being in political science...and a rational choice proponent...I am in favor of theory, then model, then statitical evaluation....but unlike many of my colleagues, I am not on jihad about this.....

If anyone wanted to be truthful about statistics...than they should admit that much of what is done is pure bullshit....hence the phrase lies, damn lies and statistics....also, as one statistician said:

"If you torture the data long enough...you can get it to say whatever you want it too..."

My greatest concern is that one should examine...is there a clear question, hypothesis, are the variables selected measuring what we are looking for, does the sampling distribution model our assumptions correctly, and have we correctly interpreted our results...

For TIA: you have inadvertently stummbled upon a century old debate within the social sciences and it has not been a pleasant discussion...so don't take it personally...you a perfectly within your rights to observe behavior and call for a further exmaination based upon these findings....

Much of the problem lies with those fields of social research, specifically economics and political science who try to be like the natural sciences, who have become virtual Pharacies of statistics...holders of the true science....as a proponent of rational choice, we ran into this alot...it is probably the most misunderstood model of human behavior out there....

[dsc]

43. Here is what many of us have a problem with

I see stats misused all the time to the detriment of left wing ideas. I vividly remember watching Crossfire one night in the late 1980's and seeing a man from the religious right saying he had proven that gay men and lesbians die at ages significantly younger than their straight conterparts. He had even taken out AIDS deaths. What did he do? He collected obits from gay and lesbian papers, threw out the AIDS deaths and calculated an average age. The problem? Stonewall happened in 1969 and thus many gays and lesbians who would have been say 75 in 1988 would have been 56 in 1969. Thus they quite likely would have been closeted gays and lesbians. He wound up asking at what age do young gays and lesbians who die, die. That would be young.

Frankly I see more than a little of his in many of these analyses. I didn't read the Kennedy stuff but the election stuff I did and commented on alot.

[Jackpine Radical]

16. Nuts.

It's enough to show a conisstent relaitonship between two variables to have a scientific finding. It's predictability, rather than cause and effect, that researchers hang their hats on. For example, for many years nobody knew how aspirin reduced headache pain, but nobody doubted that it does, because the event of taking an aspirin was followed by the event of reporting less pain on a fairly consistent basis (although not EVERY time). One could test against placebo, for example, and show that pain reductions are reported a significantly higher percentage of the time with aspirin than with placebo, and still not know why aspirin works.

[treepig]

17. predictability can only be achieved through a cause and effect relationshi

and even then you must use caution!

first, regarding your aspirin example, it is not a scientific finding to make a statistical correlation between taking aspirin, and subsequent reduced headache pain. but, as you suggest, if you then use this information to lay out a hypothesis stating that aspirin reduces headache pain and do the placebo controlled studies you mention, then you have a scientific finding.

so when TIA uses his impeccable statistics to predict the outcome of future elections (instead of cherry picking results afterwards), then he'll have moved beyond the entertainment stage.

one more example about statistics - in the 80's i established a statistical correlation between the age of my sister's turtle and the distance of the voyager 1 spacecraft from earth. there was a very consistent relationship between the two variables; in fact the correlation co-efficient was greater than .99999! i even derived an equation that could predict one variable based on the value of the other. according to your statement "It's enough to show a conisstent relaitonship between two variables to have a scientific finding. It's predictability, rather than cause and effect, that researchers hang their hats on." i had come up with a rock-solid scientific finding. common sense would indicate otherwise, however, since there is absolutely no cause and effect relationship between this statistically unasailable finding.

[JNelson6563]

10. you know what I like?

I like averages. Like Al Franken I think they are fun to play with. As he writes in Lies and the Lying Liars That Tell Them, you and Bill Gates have an average net worth of however-many-billion dollars.

Julie

[blm]

8. If they can't appreciate what you do

and the sincerity that you bring to your posts, well...that's THEIR problem.

You be you and that's fine for most of us.

;)

[Generic Other]

13. Having been called a McCarthyite Klansman this morning on DU myself

I don't know if you want any sympathy from me.

I would have defended you but not about math! You didn't want my help.

I agree with you that the odds of coming up with that number have to be pretty high. On the other hand, maybe it's some unnatural computer glitch and I should take advantage of it by playing that alchemist's number from now on in the lotto hoping to accidentally trip it.

What are the odds I would win eventually if I buy a ticket every week?

[Eloriel]

20. I don't like math

and it doesn't like me back.

HOWEVER, I have enough common sense to know that some of the arguments you've been getting on this thread are plain ole ignorance, no matter how "good" they may sound to some. Pure ignorance.

I respect your work, and your efforts, even if I don't understand them (and don't much want to, which you shouldn't take personally for the reason stated in the subject of my post). NO ONE can tell me there's not some way to show, using statistics, that three Repugs wiunning by exactly 18,181 votes in one county in Texas isn't some sort of anomaly. No one. So I'm 100% on your side.

Eloriel

[TomNickell]

22. a priori vs a posteriori

TIA,
I looked at your Kennedy Assasination post.

The difficulty with the argument is that you are, in effect, making the bet -after- the cards are on the table. Define any 4 cards ahead of time and the odds of getting -those- 4 are exactly the same as the odds of getting 4 aces. It's just that '4 aces' is pre-defined.

In statistics texts there are discussions of 'pre-planned comparisons'. If you announce your comparisons -before- you do the statistics, you can do a straight probability test. But, you can't just run hundreds of tests and look for the ones that come up statistically significant.

You (or somebody) have found 15 'witnesses' who died "mysterious" deaths. But the defiinitions are subjective and made after-the-fact. The suspicion is that you have looked through lots of possible comparisons and found the ones that look good.

In addition, the people involved were living pretty shady lives. Hence, the chances that they, or their associates, would die under unexplained circumstances is much higher than that of (say) your average suburbanite.

A possible test would be to find some random murder victim of similar circumstances to Jack Ruby and put in a similar amount of effort in looking for 'mysterious' deaths of associates. Nearly impossible, of course, but that's what would be necessary.

[TruthIsAll]

26. Tom, your argument ain't worth a plumb Nickel

You say:
I looked at your Kennedy Assasination post.

The difficulty with the argument is that you are, in effect, making the bet -after- the cards are on the table. Define any 4 cards ahead of time and the odds of getting -those- 4 are exactly the same as the odds of getting 4 aces. It's just that '4 aces' is pre-defined.

I say:
Huh? That's like saying that the chances of 15 high schoolers from a graduating class of 1000 meeting violent deaths one year after graduation is the same as 15 JFK witnesses dying the same way one year after the murder from a Warren Commission list of 1000 witnesses.

You make the same bogus argument of a priori vs. post priori. Of course its post. We are looking at an HISTORICAL event aren't we? The point is, were the deaths mentioned at all in the Warren and HSCA reports? If so how many? If not, why not?

You say:
In statistics texts there are discussions of 'pre-planned comparisons'. If you announce your comparisons -before- you do the statistics, you can do a straight probability test. But, you can't just run hundreds of tests and look for the ones that come up statistically significant.

You (or somebody) have found 15 'witnesses' who died "mysterious" deaths. But the defiinitions are subjective and made after-the-fact. The suspicion is that you have looked through lots of possible comparisons and found the ones that look good.

I Say:
That's a bogus strawman. You're very good at that, Tom. These deaths are fully documented in Jim Marrs "Crossfire". No cherry-picking here. By the way, he left out Tippet and Oswald. Also, in case you are unaware, heart attacks have been known to be induced, but I assumed these were natural deaths (give you the benefit of the doubt).

You say:
In addition, the people involved were living pretty shady lives. Hence, the chances that they, or their associates, would die under unexplained circumstances is much higher than that of (say) your average suburbanite.

I say:
Nonsense. These were INVESTIGATIVE reporters, COPS, night clubbers, etc...mostly in their 30s and 40s. They all had one thing in common. They had information which could not be revealed. Yet, otherwise, they were just Americans, like you and me.

You say:
A possible test would be to find some random murder victim of similar circumstances to Jack Ruby and put in a similar amount of effort in looking for 'mysterious' deaths of associates. Nearly impossible, of course, but that's what would be necessary.

I say:
Sorry, Tom. I'm not going to do an impossible investigation for you. Just look at the facts and circumstances of those deaths already documented to have been related to the assassination, and calculate the probabilties of them meeting violent deaths in 1964 (Warren Commssision) and 1977 (House Select Committee on Assassinations).

In EACH of these years, the odds are more than astronomical that these deaths were NOT due to coincidence:

It is Thousands of Trillions to One, ASSUMING 1000 WITNESSES.

One final caveat: This is a purely MATHEMATICAL problem from Probability 101.

The ONLY question is: HOW MANY WITNESSES ARE REFERENCED IN THE WARREN COMMISSION AND HSCA REPORTS AND HOW MANY OF THEIR STRANGE DEATHS ARE MENTIONED?

If 1000 and 15, respectively, the results stand for themselves.

If there are more than 1000 witnesses, the probability of foul play decreases. Check my post. But even if there are 5000 witnesses, the odds are still astronomical.

AND DON'T FORGET IT HAPPENED TWICE (IN '64 AND '77). THE ODDS ARE MUCH HIGHER THAN PREVIOUSLY STATED.

JUST MULTIPLY THE PROBABILITIES:

<<<<<<<<THOUSANDS OF TRILLIONS * THOUSAND OF TRILLIONS TO 1>>>>

IT'S TRULY MIND-BOGGLING....IF THOSE NUMBERS DON'T GRAB YOU, NOTHING WILL. ANYONE WHO STILL FEELS OSWALD DID IT ALONE IS JUST TOTALLY IGNORANT OR IN DEEP DENIAL..

Tom, try as you might, you are not going to draw me in to an extended argument. I have seen your battles with STICKDOG. I have better things to do than to try to knock down an infinite army of strawmen.

[TomNickell]

29. Don't think we'll get anywhere but I'll try....

1. 15 of 1000 (1.5%) doesn't sound like a real high mortality to me. These are -not- a random sample; they are selected because they had some relation to the assassination. "INVESTGATIVE reporters, cops, night clubbers, etc" aren't the same population as 'housewives, teachers, preachers'. All the people on the witness lists had -some- connection to the assasination. Or they were 'Investigative Reporters' making their living in the Conspiracy Theory Industry. NOT typical citizens. A useful comparison might be the mortality of the graduating class of some inner-city school. You -could- do an investigation like that.

If the sample is not random, the statistics are meaningless.

2. Pre-planned comparisons. There are excess deaths in the list of witnesses in Warren commission report. How many -other- lists could have been used? People in Dealy Plaza, Acquaintances of Jack Ruby, Acquaintances of Lee Harvey Oswald, Dallas cops.... What's the probability that any -one- of the possible lists had an excess mortality purely by chance?

And, how many -other- lists of 'unusual circumstances' might have been tested before finding 'excess violent deaths'?

3. It is -not- purely mathematical. That's the very easy part. You have think very carefully about the samples and the validity of the comparisons.

4. What's a 'Strange Death', or 'Violent Death'? Was this defined ahead of time?

5. Yes, it's historical. That -may- mean you aren't able to make the statistical argument.

6. Exactly what would be the point of killing off these people -after- they have testified? Wouldn't it have been more effective to get them beforehand?

These are entirely reasonable questions, which you would have to answer if you were trying to make a statistical argument for an academic publication. Maybe you can do that, but you have to take them seriously if -you- want to be taken seriously.

[TruthIsAll]

32. I said I wouldn't be drawn in, but I can't resist..

You say:
1. 15 of 1000 (1.5%) doesn't sound like a real high mortality to me. These are -not- a random sample; they are selected because they had some relation to the assassination. "INVESTGATIVE reporters, cops, night clubbers, etc" aren't the same population as 'housewives, teachers, preachers'. A useful comparison would be the mortality of the graduating class of some inner-city school.

If the sample is not random, the statistics are meaningless.

I say:
You are completely ignorant of the underlying mathematics here. This is above your level of understanding. If you had a course in basic probability and followed the full analysis, you would NOT make the statement that 15 out of 1000 dying in a given year is no big deal. How many of your high school graduates died in the year after graduating? I assume they were not drafted into Nam. Can you name one?

Random sample of what? The death rates used here were based on the full population. Of course those here were selected because they had some relation to the assassination. This is just basic. I really should just stop here. Boy, Tom, you're good. You can make statements that are completely bogus and make them sound legitimate. THIS ONE STATEMENT OF YOURS IS SO OFF THE WALL, SHOWS SUCH A LACK OF BASIC STATISTICAL ANALYSIS, THAT I SHOULD REALLY JUST STOP RIGHT HERE.

You say:
2. It is -not- purely mathematical. You have think very carefully about the samples and the validity of the comparisons.

I say:
You've got to be kidding. It is purely mathematical just like the work of an actuary in predicting mortality rates is purely mathematical. Tom, we are dealing with a finite universe of individuals here (witnesses to the assassination). We analyze their mortality rates as a group. When 15 out of 1000 from ANY group die in a given year, we have a very suspicious occurrence. In fact, I defy you from finding such an occurrence. It has never happened. Except if you were in a a marine regiment in Nam. Or a witness to the JFK murder.

You say:
3. What's a 'Strange Death'? Was this defined ahead of time?
I say:
Read the post: A strange death is murder, accident, suicide or some unknown factor.

4. Yes, it's historical. That may mean you aren't able to make the argument.

I say:
What BULLSHIT!

You say:
Overall, you just can't make very strong inferences with this kind of argument.

I say:
Why do you do this? Have you no other interest than in just disrupting legitimate threads here at DU? Are you really that OBTUSE? I don't believe you are. I think you know exactly what your doing. But to what purpose? Do you really think you will change any hearts and minds with your bogus arguments? Or is that not your intent? Or do you just despise those of us who don't believe in coincidence theories as you do?

[TomNickell]

35. Off the Wall?

I did a big edit on the original post, you might want to go back and look at that.

<<Random sample of what? All 150-200 million Americans livinfg at the time? Of course they were selected because they had some relation to the assassination. This is basic. I really should just stop here. Boy, Tom, you're good. You can make statements that are completely bogus and make them sound legitimate. THIS ONE STATEMENT OF YOURS IS SO OFF THE WALL, SHOWS SUCH A LACK OF BASIC STATISTICAL ANALYSIS, THAT I SHOULD REALLY JUST STOP RIGHT HERE.>>

What's off the wall?
Actuarial tables are drawn from the whole population--250 million or so. But, subsets are going to have completely different statistics--and different life-insurance rates. Diabetics, Vietnam vets, Marine regiment in Nam, Smokers, Race Car Drivers, Drinkers, Residents of downtown Dallas or New Orleans, Investigative reporters, Ex-convicts, Free-lance writers, Conspiracy Theorists--all have different mortality rates. And rates of violent death. It is absolutely necessary to make a comparison to a comparable group. 15 of 1000 might well be excessive, but you need to show some comparison.

If 1.5% of -any- group have never died in a given year, produce the data. Or, more reasonably, select some high-mortality group and find their mortality statistics. Ex-convicts?

<< A strange death is by murder, accident, suicide or some unknown factor.
>>
That really leaves a lot of wiggle-room. What's an 'unknown factor'?

And shouldn't some of the murders and suicides be excluded because the circumstances were clear?

<<Have you no other interest than in just disrupting legitimate threads here at DU? Are you that OBTUSE? >>

In what way am I disrupting this thread? I thought I was quite polite, actually.

I am asking sensible questions. Perhaps my arguments -sound- sensible because they -are- sensible.

You would have to answer identical questions if you were writing for an academic course or publication.

-Some- threads around here are so loony that sensible questions are just not possible. Wild theories built upon no evidence at all. Or logic twisted into incomprehensible tangles. I post my opinions of these theories. It will not convince the True Believers, but it may put a little drop of doubt in their minds. And it may reassure some sensible folks that their Common Sense is still quite valid.

[dsc]

30. cops and nightclub owners

are not like you and me when it comes to the odds of dying a violent death. I would agree that shady was a poor choice of words on the poster's part but certainly both of those classes of people are vastly more likely to die violently than say a school teacher, doctor, or computer programer to name three.

[TomNickell]

36. Shady? A nice 1964 kind of word.

I wouldn't use 'shady' in normal conversation.

But lots of these folks were living in the less affluent parts of the Big City.

[Skittles]

23. WHO IS BASHING MY BELOVED TIA

I'LL KICK THEIR ASS

[TruthIsAll]

25. Thanks, Skittles, but this GUY can take care of himself...

I guess I write like a woman..

[Skittles]

44. Naw, I don't discriminate based on gender

I just don't like the idea of people attacking you. Many times I have done a search on your posts during conference calls. Great stuff~!

[MODemocrat]

33. This is intriguing and is worthy of research

I was young at the time of JFK's death, but the one thing that stuck in my head was the timing of the gossip columnist, Dorothy Kilgallen's
death. She had said that she was going to "Bust the Kennedy Assassination wide open" but, ironically, she died in her sleep, and the coroner couldn't confirm whether it was a suicide or murder.
She had interviewed Jack Ruby, prior to her statement.
I did a google search today using Dorothy Kilgallen (columnist), and it contained quite a few things that I never knew.

[nwstrn]

34. I love junk science!

What a hoot! This is a fascinating topic. Since you "welcome" "criticism," I have a question: Do the 1999 statistics give the probability of death by bees? Say two members of a 10-person family die from bee stings, in separate incidents and cities, on the same day. Astronomical odds, right? (assuming we had not been taken over by killer bees). You could go through the same analysis set forth above, and come up with the same conclusion: "not too likely that these deaths" were from bee stings. Does this "prove" anything? Would your research "prove" that the events did not happen--i.e. that the people did not die from bee stings? No. A past event cannot be given a "probability" that proves or disproves the occurrence of the event itself. Nice try, though.