Hey Fire, you speak Math: What does this mean?

[TayTay]

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x351187

This is the link to the report on statistical anomalies in last fall's election. What doe sthis mean? (I don't speak statistics.)

Thanks!

[Firespirit]

1. I'm going to take a look at it later this afternoon

I really shouldn't be on DU at all right now, should be doing something else, but here I am.... I'll read over this and see what I can make of it. Statistics aren't a foreign language to me. :P

I heard that BS excuse that the Bushies were somehow less talkative and it never seemed plausible at all. What I hope this report addresses is whether the exit poll was a statistically valid random sample.

[TayTay]

2. Thank you so much!

In turn, someday I will look up an archived item for you. (Boy, is that lame or what?) Okay, name your price --- Pics, stories that mention JK & sexy in the same sentence, whatever you want.

You da greatest, kiddo!

[Firespirit]

8. And now my "expert" opinion, at last.

:P

This was a very interesting paper. I will try to paraphrase what I found to be the most striking points:

1. The poll is proven to be a random sample. For a random sample, the errors (or discrepancy between the poll results and the reported results) for the 50 states would conform, more or less, to a normal distribution or "bell curve." The paper shows that this DOES happen, making it a random sample and sound reweighting practices.

The only thing is, the median for this curve is NOT zero error, as it should be if the polls and reported results lined up. The median is shifted to the left, which with their notation means that the error (or discrepancy) for the polls favored Kerry.

2. The paper demonstrates that the "reticent Bush voter" idea is pretty much absurd and offers up STRONG circumstantial evidence for the "padding of Bush votes in Republican areas" idea that has been advanced before. The pollsters had hypothesized that Bush voters were reluctant to take the poll in mixed political company or in predominantly Kerry precincts. If that were true, then the error rates for those precincts -- the differences between the polled numbers and the reported tallies -- would be greater than in predominantly GOP areas. However, the greatest errors were found in heavily GOP precincts.

They also found that in order for the "reticent Bush voter" idea to be true, not only would it require that Bush voters kept quiet in their own strongholds, but that Kerry voters were loudest and most willing to be polled in the strongest Republican areas and least willing to be polled in their strongholds.

3. The paper found that the polls, taken as a national average, were way off in areas that counted votes mechanically or electronically as opposed to areas that counted them by hand.

4. The paper mentions that the poll for Senate races was well within the margin of error and overall predicted the Senate races accurately. If the "reticent Bush voter" idea were true, then it would have carried over and skewed the Senate race poll results as well, because the exact same people were polled for both contests. But that didn't happen.


All in all, quite interesting.

Apparently the paper has been vetted by lots of professional statisticians and mathematicians and has stood up, BTW.

[TayTay]

9. I don't understand this part

The only thing is, the median for this curve is NOT zero error, as it should be if the polls and reported results lined up. The median is shifted to the left, which with their notation means that the error (or discrepancy) for the polls.

Does this mean that the exit polls skewed to Kerry?

[Firespirit]

11. Essentially, yes.

The error rate, or discrepancy if you prefer (as I do, because I don't want to say that the polls are in error), was in Kerry's direction in the polls. That is to say, with a fair and accurate count, the polls would have been biased in his favor.

[ginnyinWI]

10. thanks for making it make sense in English!

I too, glaze over at seeing a page full of stats. I got as far as college Algebra, but not stats.

So the exit polls were accurate for Senatorial races, and not for the presidential? Easiest way to explain that is votes added by tabulator.

[whometense]

12. I haven't any expertise at all,

but what you say sounds logical. I still feel in my gut that there was tampering. Not to sound all :tinfoilhat:, but I can't forget about the trial run with Max Cleland.

Do you know what the status is of the legal case in Ohio? Do you think that will go anywhere?

[sandnsea]

7. On these numbers

I'm not a math person either. One thing that has stuck with me, Oregon kept its Dem & Nader voters, percentages are the same between 2000 & 2004. We're one of the only states that did, if not the only. Even CA lost a percentage of Dem/Nader voters compared to 2000. How can that be?

We have all paper ballots and the paper IS the ballot. They are mailed to the county and all ballots counted at one county location. Then they are transmitted to Salem. No precinct uploads.

My gut keeps telling me that looking at the differences between Oregon and other states will either show where we lost votes, or show where votes were stolen. There is no reason Oregon would respond any differently than Ohio. We had gay marriage on the ballot and have our share of fundies and militia types. We also land use deregulation, forest practices, and marijuana initiative. Plenty of stuff to bring the wingers and conservatives out to vote.

What we don't have is an easily manipulated voting process. Or a high minority population like OH or FL. We also have a Dem SOS. I just keep thinking that comparing voting patterns could identify irregularities. We do have precincts, but we mail all ballots to the county.

Does this make any sense?

[karynnj]

3. Their analysis is compelling

Assuming that the data they used and the number crunching they did was accurate, they make a very good case that the "Kerry voters were less shy about completing the exit polls" argument does not work.

- They find the largest errors in the most heavily Bush areas.

- The same exit polls, unadjusted were accurate for the Senate and house races. To make the exit polls conform for the Presidential election, they had to make major shifts - so in the adjusted data D/R is 37%/37%, but from the data itself, the %D is higher than % R. This is interesting because the same people filled out the exits polls for President and the lower races. On an intuitive logical level this may be the strongest argument that the sample itself wasn't bad.

I really need to look at it in more detail, but assuming the data and tables are correct, I think Kerry may have won. (But statistics won't prove it beyond a reasonable doubt for many if not most people.)

[TayTay]

4. Oops, I'm sorry I didn't include others

in this analysis. I should never assume that just because my math brain is missing, that everyone's else's is. Sorry. Anyone who isn't me should comment about what they know on this.

Dumb TayTay

[karynnj]

5. I didn't take it personally

I felt I should look at it. My degree is in Applied Mathematics, not Statistics, but I worked for about 24 years in various groups that did statistical analysis (as well as other analytical stuff) for AT&T - much of which was sampling.

Their observation that the sample was good for the other races makes me very suspicious. I had been suspicious because the report really said they had found no design or mathematical problems, leaving only the data collection as a possible problem. So they simply said our sample has too few Bush voters to project a Bush win, and then HYPOTHESIZED that it was because they shied away from doing the exit polls. My problem with this is that it would have been easier to believe about Bible belt Kerry voters. (some of whom were told it was a mortal sin to vote for Kerry - I wonder how many confessions went "Bless me father for I have sinned. My last confession was ...ago. The sins I have committed are: I voted for Kerry,...." )

[TayTay]

6. Wow! I 'm impressed

You and Fire represent a real treasurehouse of expertise. This is a wonderful thing. I am going to attempt to read the report and I'm sure I will have about a zillion questions.

I bet this is being read by a whole lot of former Kerry staffers and by Sen. Kerry's family. I believe Kerry's daughters have both made public comments about Ohio and regard the results with outright suspicion.

Again, thanks for helping me make sense of this. I long ago came to terms with having no math brains (or spelling brains either.) I have learned to compensate (LOL! I am joking, of course.)

[Firespirit]

14. It *would* have been easier to believe about Bible Belt Kerry voters.

Like my state, for instance, which is most definitely Bible Belt -- and which I suspect may have been "padded" for Bush because the exit polls were off and many areas of the state use Diebold OptiScan, which was one of the technologies that fell under suspicion.

The polls were giving Kerry 43-45% of my state's vote. He reportedly got 39%. That discrepancy would suggest that even in a majority-Republican state, this same "reticent Bush voter" pattern took place, and that the Kerry voters were more willing to take the poll. I could tell you from anecdotal experience that the Bush voters I've met were NOT shy at all about their opinions. In fact, they assumed that anyone who was Caucasian and middle-class-looking had to be a Bush voter, and if they talked politics with such people, their assumption would be quite obvious from what they said. (A few who didn't know me well actually made that assumption about ME.) The Kerry voters I met weren't ashamed or afraid, but they didn't make such unjustified assumptions either, and there is simply no way that they intimidated the Bush voters.

[Firespirit]

15. Another problem with the hypothesis...

...is that the polls were off, on an average, nationwide. Sure it was worse in some areas than others, and some polls were actually fairly accurate. I find it VERY hard to swallow that this pattern could have taken place all over the nation. One region -- perhaps. I could buy it that Bush voters would be shy in, say, New York City, or in San Francisco. (Or Boston, for that matter.) But not nationwide.

[karynnj]

17. Good point,

Your summary was really nicely done. I agree with you that the nationwide aspect is hard to swallow. Also as this skew in the liklihood of people being willing to fill out the exit polls did not exist in previous elections, there needs to be a reason for why people would be afraid or embarrased for their vote to be known.

What I haven't seen is a description of the actual sample design or, lacking that, a statement that the design was similar to those used in past elections. The exit polling company's shy Republican explanation is clearly not credible and it is desturbing that it is self serving. They are saying that their sample design, implementation and analysis were correct - but there was this data collection problem THAT NEVER WAS A PROBLEM BEFORE that caused the results to be significantly wrong.

I would believe that you could have this kind of skew, if there were a minority or a woman running. People, who on the issues, would be predisposed to vote for a candidate of that candidate's party might be ashamed that they voted for the opponent because of sexism or racism.

[TayTay]

13. I have another question for anyone who can answer it

Kerry got 6 million more votes in 2004 than Gore did in 2000. Does anyone know if there is a web site that collects where he got more voters and among what groups? I am willing to 'do my own research' but I was just wondering if anyone knew if there was a site with this info before I set off on a quest to find it.

Thanks!

[Firespirit]

16. And it gets even more suspect on a second reading.

I just re-read this thing, and... well... HOLY CRAP! This report sends up a HUGE red flag. Statistics aren't proof, but we're getting damn close to it now.

----

A poll should be a random sample representative of the whole data set. The measurement error for a random sample is normally distributed, and for our sample -- comparing a poll to a dataset of votes -- should be clustered around zero, with deviations equally likely to have occurred in both directions (i.e., errors favoring Bush and Kerry should have been equally likely). Sort of like this:



Non-normal measurement errors indicate a problem with the sample. If the poll's sample collection were bad, and Mitofsky had failed to draw a random sample and weight it properly, then the measurement error would not have been normally distributed. It would have been lopsided:



Now here is what OUR measurement error curve looks like:



We have a normally distributed measurement error, as the paper shows, but it is not clustered around zero. It's shifted left.

I need to do some more research on this, but intuitively it would seem to suggest that if we have a normal measurement error distribution that is centered around some number other than zero, and we're expecting it to be centered around zero, there is a problem with the dataset. Especially if our sample (the poll) is not a set of values drawn EXPLICITLY from the dataset.

See, ordinarily, you'd take a random sampling of measurements from a very large dataset by directly measuring your variable. The poll technically doesn't do this, because whether one believes in fraud or not, asking 12,000 people how they voted is NOT the same thing as directly observing a random sample of 12,000 ballots being scanned and tallied. It's not really sampling the same dataset. The dataset of all tallied votes (let's call it Set T) and the dataset of all 115 million voters' responses to a poll (which we'll call Set P), had such a thing been taken, are not the same. If we can have confidence in the counting, then they should represent the same data, but they aren't the same collection. The poll is sampling one and claiming that it represents both.

Clearly, this poll did not represent the set of tabulated votes--Set T--but it was normally distributed, which means it was a random sample of Set P. That implies that Set P was shifted one standard deviation away from Set T. So again, we either have this bizarre nationwide pattern of Bush voters being quiet (or maybe even lying about how they voted), or we have a problem with the counting. The paper offers up very good evidence against the idea of Bush voters being quiet or lying about their votes. The biggest errors favoring Bush occurred in strongly Republican areas where they would not have any reason to feel embarrassed about their votes, had they voted for Bush.

//Edit final name of that paragraph. Typo. Bush, not Kerry. :P


---
Sorry for the long nerdy post. I hope I haven't confused anyone.

[TayTay]

18. Not nerdy at all.

This is exactly what I asked for. You and karrnnj are doing great. Your explanations are clear and concise. I am getting it.

And thanks to both of you. Much appreciated.

[TayTay]

19. Hey read this: (found via DKos)

http://itmanagement.earthweb.com/columns/executive_tech/article.php/3495176

I think this just might get out of the indy media. (then again, I was late giving up my belief in Santa and the Good Fairy. Sigh!)