Another Mathematician Signs- Math Proof of Bunk of Mitofsky Analysis

[sunshinekathy]

Two more people have contributed or reviewed and signed on to the:

Mathematical Proof that Election Sciences Institute's Test to Rule Out Vote Fraud Is Logically Incorrect

http://electionarchive.org/ucvAnalysis/US/exit-polls/ESI/ESI-hypothesis-illogical.pdf

Vickie Lovegren, Ohio Case Western math instructor, suggested great edits to the display of my counterexample so that folks will understand better why the Mitofsky/ESI analysis is all bunk.

Even Better, Vickie added an entire new symbolic math logic proof based on my counterexample that I added to the appendix because I don't have time to integrate it fully into the paper.

In fact I REALLY NEED HELP promoting this paper to the press because I do not have any time to call press and leave messages with their news desk reporters and email them personally.

PLEASE HELP by seeing that this math logic proof gets into the hands of press if you have any time - especially in Ohio.

Another PhD (physicist who calls mathematics pathematics) also reviewed and added a short paragraph, who has also an excellent academic background and is a retired teacher/professor(not sure what his title was) from Illinois Institute of Technology.

For those who complained earlier that the math proof didn't look enough like a proof, Appendix A will satisfy the most stringent I-want-it-to-look-like-a-proof people.

If you pass it around, please use this latest version:

http://electionarchive.org/ucvAnalysis/US/exit-polls/ESI/ESI-hypothesis-illogical.pdf

[texpatriot2004]

1. 2 more join the proof. Yeah. Congrats. Thanks too. n.t

[Psephos]

2. Bad juju

Politics and science. Nary the twain shall meet...except at dire peril to the truth.

Seriously, history is an endless litany of evidence that if you have a political view, you can and will find plenty of evidence and experts to support it. More to the point, you will not be able to see any merit in evidence and experts that don't support it.

I'm not taking a position yea or nay on the Mitofsky/ESI analysis, and I don't intend to, either. Not one person on either side of that "debate" will convince someone from the other side, because each side drew its conclusions first, and only then adduced its evidence.

Peace.

[sunshinekathy]

4. I am not like others in that way

I make up my mind on the evidence and always strive to stay open-minded.

For instance, the exit poll data clearly shows evidence of Democrats manipulating or tampering with vote totals as well as Republicans.

OK, to be more "scientific" about that claim, let's just say that there are over-estimates of the Bush vote that are well outside the margin of any sampling error of any exit poll, as well as over-estimates of the Kerry vote.

Other people might want to claim that ONLY the Republicans are tampering with vote counts, but that would be contrary to the facts, which is what I always go by. I adopted truth as my "motto" for lack of a better word, at such a young age, that I would have trouble clinging to ANY point of view that was not factually supported.

And why would we NOT expect both Democrats and Repubs to be tampering with vote counts (and not just in the presidential race either, but also in US and State house and senate races and governor's races and who knows what else?) when we have a vote counting system that allows ANY insiders complete freedom to tamper with vote counts and never get caught - so why wouldn't they tamper when the payoffs include control of budgets from the millions at the county level to the trillions at the federal level.

Until we do two things:

1. independent audits of vote count accuracy in ALL elections and all races, and

2. election data collection and monitoring;

we are deluding ourselves if we even imagine that our elections are being counted accurately - ANY of them - not just the presidential races.

Of course, the number of precincts that look suspiciously like Kerry got too many votes is much much fewer than the number who look suspiciously like Bush votes were added.

You see, we're not all the same when it comes to being willing to look at and decide based on evidence. I have been told that a lot of folks make their minds up based on how much they like someone who has an opinion, or by how many other people believe something - not I.

Best,

Kathy

[OnTheOtherHand]

18. actually, it is (somewhat) better than you think

Behind or beneath the sound and fury, people are convincing each other of all sorts of things. But the public discourse is frustratingly polarized, and I think you are wise not to choose a "side."

Incidentally, I don't think that any participant in this debate thinks that s/he drew conclusions before adducing evidence. It isn't just the confluence of politics (in the usual sense) and science, although that certainly exacerbates the conflict. Every riproaring debate -- scientific, political, or both -- tends to leave both sides thinking, "How could anyone actually believe that?"

IMHO there's a wide range of things one could believe about the election. However, I don't see how it is permissible to believe that ESI "rule(d) out vote fraud," and so I think the new USCV paper is unacceptable from its title onward. (Unless they have changed the title by now -- I think the paper is on at least its fourth version.) I learn a lot more about possibilities of fraud from other sources.

[Febble]

3. No amount of editing

will alter the fact that you mis-state the hypothesis you are "de-bunking", as well as its proposers' conclusions.

And will you PLEASE for the umpteenth time of asking, correct the quotation from me, in Footnote 4, and give the link to it. Here it is:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=398055&mesg_id=398338

Copy and paste if you cannot see your error. The error is that you have used the word "indicates" where I used "indexes". It does make a difference, as I have pointed out to you, and failure to correct verifiably erroneous quotations when repeatedly pointed out does not inspire confidence in any of the rest of the paper. Moroever, people who can't correct their own errors prior to publishing work are in no position to upbraid others for the same perceived offence.

And of course, needless to say, I maintain your that your paper, claiming to be "Proof of Bunk" is no such thing.

But until you actually address the hypothesis referred to and the conclusions made by those who tested it, it cannot be.

Frankly, I find it "irresponsible" for mathematicians, with no apparent knowledge of statistical method and assumptions, to publish such a "misleading" paper, and highly offensive for the authors then to take to task the people whose work they characterize as "bunk" for not shooting straight.

The logical fallacy the paper makes is known as the straw man. Or one of them is.

Elizabeth Liddle.


(edited to re-place accidentally deleted word)

[sunshinekathy]

5. It doesn't matter Febble - we've proved your analysis' is meaningless

Whether or not we've stated the hypothesis exactly the way YOU want it stated (and we simply quoted it from ESI), we've proved that all the analysis that you've been doing comparing the 2000 and 2004 Bush vote share and WPD is simply wasting everyone's time and misleading them into thinking that you can make conclusions based on the data when absolutely nothing can be concluded from the type of analysis you've been doing. It's just a simple fact.

Again, take the proof AND your analysis to a math professor at the university where you are studying for your PhD and ask them to explain it to you PLEASE.

Thank You.

Kathy

[Febble]

6. No you didn't quote it directly from ESI

you added a crucial interpolation. Without the interpolation, the hypothesis is perfectly logical - self-evident, in fact.

And I could just as easily argue that you are wasting quite a lot of other people's time (including mine) by diverting attention from the very real information that we can gain from both Mitofsky's and ESI's analyses.

Luckily, I am married to a Professor with a PhD in Maths and Physics, so I won't have to bother anyone at my university. Fortunately he is also a pretty keen statistician as well.

[sunshinekathy]

7. Can you be specific?

What EXACTLY is the difference between our statement and ESI's? We quoted it exactly.

We even analyzed exactly how ESI was using their invalid inference by looking at the contrapositive and we looked at their analysis to be sure also.

So, you seem to be obfuscating and not being specific.

If you really had a valid point, you would give a specific. We KNOW we exactly represented ESI's logic because otherwise ESI and Mitofsky would not have been able to rule out vote fraud using the contrapositive.

We assumed that both ESI and Mitofsky were intelligent enough to know whether their hypothesis, if valid, was capable or not of using the contrapositive to do what they claimed it did and we also carefully looked at how they analyzed the data in their paper.

You didn't even know what they'd stated in their paper yesterday when you first began (wrongly) claiming that we misstated ESI's hypothesis.

This time you have not given anything specific to back up your claims that we misstated the hypothesis.

I would like to point out, however, that even if we did slightly misstate the ESI hypothesis, which we did not, we proved without any doubt that ESI, Mitofsky and your analysis comparing the 2000 and 2004 election results with WPD is meaningless and useless for determining anything about vote faaud.

So what is the point of continuing this silly misrepresentation that we misstated ESI's hypothesis?

[Febble]

9. I was specific countless times

on your other thread.

the interpolation is the words "than in 2000":

If systematic fraud or error in vote counting {favoring Bush} occurred {in precincts} in 2004 but not in 2000, {then} Bush would have done significantly better in those precincts in 2004 {than in 2000}, and we would see larger differences between the reported vote and exit poll in those precincts {than in other 2004 exit-polled precincts}."

It is the interpolation in bold that falsifies the statement, which you correctly "prove" is, with the interpolation, indeed false. Without the interpolation the statement is perfectly valid.

[sunshinekathy]

11. That is exactly how ESI and your own analysis uses that inference

Look at ESI's paper (which you didn't seem to know this was quoted from yesterday) and at your own scatterplots and analysis to see that you clearly subtract Bush vote share in 2000 from his share in 2004.

[Febble]

13. Of course I know the ESI paper

but I have to admit, not by heart, and didn't recognise at first you had quoted it verbatim, precisely because of your extrapolations which render it, as you point out, patent nonsense.

Yes, the measure of swing involves a subtraction in the case of the ESI analysis (and something similar in Mitofsky's). But that is not the same as saying that fraud will result in Bush necessarily going above zero (i.e. doing better than in 2000). Fraud will simply result in him doing better than he would otherwise have done. As I said, on the other thread.

You seemed to think that this was untestable. Well it isn't. You can test it with correlational analysis, as was done, and you compute the ratio between explained and unexplained variance (or shared, and unshared variance). It's how you infer a relationship between two variables. If you don't find one it doesn't mean it isn't there, rather that it is swamped by variance from other sources. But it does put major limits on the effect size, which means that you can conclude that the effect of the shared factor is small.

Which is why both ESI and Mitofsky conclude that fraud does not appear to be the cause of the exit poll discrepancy.

However, we could be cleverer than that, and try to figure out how fraud might have been the cause of the exit poll discrepancy and yet not produce a correlation. You did so in your paper. I have done so on many threads here. Which is why I rather resent your repeated accusation that what I personally have done is BUNK. Doing these exercises may generate good testable hypotheses. Mitofsky has offered to test any hypothesis Steve Freeman proposes.

The analyses are not BUNK. They tell us some very important things about the kind of fraud that is and is not compatible with the results of those analyses. If we dismiss them as BUNK we will deprive ourselves of that information, which could, in fact, tell us the mostly likely places to look for fraud, as well as what type to look for.

[sunshinekathy]

8. It doesn't matter- we've proved ESI comparison analysis is meaningless

Febble,

Try to understand. We've proven that the analysis you've been doing to compare the 2000 and 2004 election results and compare the differences with WPD is useless for analyzing whether or not there is vote fraud.

It is as simple as that. Quibbling over whether you liked exactly how we copied the quote from ESI's paper to ours and stated it in terms of math logic is also just meaningless waste of time, just like your purporting that analysis that is based on an invalid illogical hypothesis that Bush vote share differences from 2000 to 2004 and its relationship with WPD is informative - It's not.

I don't have time to waste discussing this endlessly with you, PLEASE take both your analysis and our proof to a PhD mathematician (not a person with a degree in another field who happens to do statistics) but a real live mathematician and see if you can get him/her to agree with you.

[Febble]

10. I would agree, and indeed have made the point

countless times that the analyses by Mitofsky and ESI do nothing to tell you whether there was vote fraud.

What they do indicate strongly is that fraud is not indexed by the exit poll discrepancy.

It's another of your straw men.

[sunshinekathy]

12. It does not indicate that

How on earth do you get that your analysis "indicate strongly is that fraud is not indexed by the exit poll discrepancy"?!

Your recent analysis doesn't show that.

What hypothesis are you using to now claim (and this is the first I've heard of this one) that you have indicated that fraud and WPD are not related or correlated?

Did you read the algebraic derivations I did in Ron and my paper that clearly show the effects on WPD of vote shifts from one candidate to another? You even created such a spreadsheet yourself as did Josh Mitteldorf. This is an all new claim.

Please responsibly research this before you publicly release another new hypothesis this time.

Are you going to write a new paper on this new hypothesis?

I think we've already disproved that one before you begin. Please read the latest version of the exit poll paper that Ron and I wrote. It is linked off the Exit Poll section on the home page of http://electionarchive.org

[Febble]

14. Well I have answered your first question

many times on DU, so you might like to search my threads.

And I really don't have time to repeat it now as I have to go to work.

But try this one, and read the comments. There are some good ones, especially by eomer, BillBored, TimeForChange, Land Shark, OTOH and others.

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=398267&mesg_id=398267

And look at Mark Lindeman's piece here:

http://inside.bard.edu/~lindeman/slides.html

[Febble]

15. For example

if fraud, as you suggest, and as others have suggested, was concentrated in large precincts, it could have had a large effect on Bush's vote, but a small effect on the exit polls.

The other possibility is that fraud may have been targetted only in precincts where Bush was doing badly, as, again, you suggest. However, I have been working with this hypothesis for some time, and so far have found that if this was done at a level not to impact on that correlation, it would also have a fairly minor impact on the net exit poll discrepancy.

However, it might nonetheless have been an important factor in Bush's win.

As might various forms of fraud that do not show up in exit polls.

But what this means is that fraud is not indexed by the exit poll discrepancy. Not that there was no fraud.

[Febble]

16. And I would suggest to you

that you take your work to a real live statistician (not someone with a degree in another field who happens to do statistics) and have them check whether your counter-example invalidates the (correctly stated) hypothesis.

Because of course it doesn't. Any correlational analyses will have countless "counter-examples". It's called residual variance. What we need to know is whether the residual variance out-swamps the model. In the case of both the ESI analysis and the Mitofsky analysis it does.


(edited for error)

[Febble]

17. Your paper's title

is also misleading:

Mathematical Proof that Election Sciences Institute's Test to Rule Out Vote Fraud Is Logically Incorrect.

(my bold)

To my knowledge, neither Mitofsky nor ESI, nor indeed I (whom I mention merely because I am cited in the paper) have claimed that the hypothesis they tested was a "Test to Rule out Vote Fraud".

ESI concluded that:

Our study indicates that the non-response rate theory is much more likely than the fraud accusation theory to account for most, if not all, of the observed discrepancy between the exit polls and the actual result.

(my bold)

Also that:

To rule out important errors or irregularities within the election system, further analysis is needed.

So the basic tenet of your paper would itself seem to be, er, flawed.

(edit: text formatted for clarity)

[papau]

19. Fraud was not selective - but all machines of a certain type were

"fixed" to report a bias toward Bush is my choice - and data in Ohio to be released may shed some light..

A lot of us have proposed the above.

Indeed with proper programing the "making random" of the above bias gives us a nicely "hidden" theft of an election.

Since all the work needed can be done at time of manufacturing, with evidence wipe out via software upgrade, Stat proof is all we can hope for until the programmer or senior management confesses.

And I am sorry, but "Test to Rule Out Vote Fraud" and replace with a "massive change in shyness of the GOP voter" concept is indeed what ESI/Mitofsky proposed - and it is indeed "Logically Incorrect" -

That these folks took the Mitofsky idea seriously is interesting - but they have killed the subtraction with 2000 idea as a test of when one multiplies by weighting ("real" votes in one case and exit poll weighting in the other) might solve something and indeed might justify the exit poll being dismissed as proof of fraud.

Afraid we still have an exit poll that screams fraud.

[Febble]

20. Mitofsky certainly

proposed that the exit poll discrepancy might have been due to lower participation rates by Bush voters.

However, this hypothesis is perfectly compatible with fraud i.e. evidence that the discrepancy was not due to fraud (which the analyses suggest) does not "rule out vote fraud" in the election. It merely (and strongly) suggest that fraud was not the cause of the discrepancy.

My own belief is that the exit poll evidence is a complete red herring, and that the discrepancy was largely due to polling factors.

However, because that does not rule out fraud, I still take seriously the many indications that the election was not honest. It could well have been stolen. But there are many ways to steal an election that will not show up in the exit polls.

[papau]

21. "strongly suggest that fraud was not the cause" only if you buy shyness

change - which I do not.

"does not "rule out vote fraud" in the election" is something we agree on - but the media has used it as if it does indeed rule out fraud.

"married to Math phd" !!!- NOW THAT IS A LADY THAT MARRIED FOR LOVE!

:toast:

:-)

[Febble]

22. Well I married him

for the physics. I thought it would be cool to be able to have someone to remind me how fusion worked whenever I forgot.

Actually he then became an MD and does research into schizophrenia using brain imaging, so he really is no slouch on the stats front. In fact he was responsible for the development of Statistical Parametric Mapping (SPM) one of the most widely used brain imaging software packages.

But hey, that doesn't mean Bush didn't steal the election....

I don't necessarily buy the "shy" thing either, but I do think there is evidence (maybe not enough to convince you) that non-response bias was a factor.

My point, made too often, but I'll make it again as people so often misunderstand the point:

I do not think that the exit poll discrepancy is evidence that the election was stolen.

However, neither do I not think it rules it out.

What I mean is that the magnitude of the discrepancy is not a measure of fraud.

Thassall.


(edit for typo)

[Moderator]

23. Lock

This is the third time that the mods have locked similar threads on this subject.

PLEASE do not continue opening threads on this subject.

Thank you