Somebody explain margin of error to me

[Hippo_Tron]

I was reading the thread about the Salon article and the exit polls and I didn't quite understand this.

Let's say that there's a poll that shows Kerry 53% Bush 47% with a 4% margin of error.

Now here's what I don't get, is Kerry winning outside of the margin of error or is he still wtihin the margin of error.

Specifically, does a 4% margin of error mean that the worst possible scenario is that the poll changes 4 points total (2 points are subtracted from Kerry and 2 points are added to Bush) or is the worst possible scenario that 4 points are subtracted from Kerry and added to Bush.

[LightningFlash]

1. I can explain it to you...

It's basically a load of nonsense. Their "last" exscuse is that Mitofsky's pollsters were ordered to seek out Kerry voters more often and went hunting them, the Democratic Exuberant Responder theory....En Masse....(Psychic interviews!)

It doesn't stack up to even a door stop. But hey, it tries its best to explain the stolen votes away. :shrug: :shrug:

[Hippo_Tron]

2. Okay, but that doesn't answer my question

I'm not asking whether the article was right or wrong, I'm asking how margin of error works.

[Internut]

3. If you want to know the full story -

http://en.wikipedia.org/wiki/Margin_of_error

[0rganism]

4. Simply put, yes it's within the margin of error

Statistics are usually reported with the "95% confidence interval", which means 95% of the time the results will be within the given range. So if the poll you saw had Kerry 53% Bush 47% +/- 4% (which indicates a relatively small sample size) then the surveyer is 95% sure the outcome is 49% < Kerry < 57%, 43% < Bush < 51%.

[Hippo_Tron]

8. Okay makes sense, so for the record...

4% means that it is...

49% < Kerry < 57%, 43% < Bush < 51%

AND NOT

51% < Kerry < 55%, 45% < Bush < 49%

[0rganism]

9. Sort of, but it's not exactly that simple

If you really want to understand it, read the wiki link (which even uses a Bush/Kerry/Nader example from Newsweek). What I stated is a raw, and even misleading, interpretation of the statistic, as it doesn't account for the actual distribution within the range, which is not even. For one thing, results close to the polled result are more likely than those at the fringe, with the standard Bayssian distribution. So Kerry 51 Bush 49 is more likely than Kerry 49 Bush 51.

[Hippo_Tron]

10. BUT, Kerry 49 Bush 51 is POSSIBLE within the 95% certainty

And then of course there is that 5% uncertainty which I guess really messes things up.

[0rganism]

11. Right. It gets messy very quickly.

And then there's questions about whether the population samples were taken in proportion appropriate to the demographics of Americans, the demographics of registered voters, the demographics of "likely" voters (now there's a whole other pickle), or the demographics of people more likely to willingly answer a survey. So responsible pollsters try to weight their polls and report their models accordingly, by asking other questions to establish the population segments that fit their models.

[Gman]

5. The margin of error depends, statistically, on the size of the poll

the more people sampled, the smaller the MOE. You can calculate the MOE based on the size of the "sample" versus the size of the "population" with the population being every eligible voter. The sample should reflect in every way the demographic make up of the population. However, it's too much statistics to explain after this many drinks tonight!

[punpirate]

6. Very simply, margin of error...

... depends upon the sample size. Small samples result in large margins of error, large samples produce a small MOE. It also has to do with a statistical norm known as the "confidence interval." That means that if you asked a question from this poll 100 times, 95 of those times, the percentage of people giving a particular answer would be within 4 points of the percentage who gave that same answer earlier.

The change could be a total of a 4-point swing.

[Selatius]

7. Self-delete

[katsy]

12. now that the good people here...

have explained it so well to you, let me give you my take on the margin of error:

It's the hole smirking chimp has never and will never ever hope to get out of even if they coronate him fucking king. He's a loser.