New Math PROOF that ESI/Mitofsky Logic is Invalid (Not Correct)

[sunshinekathy]

In two days, on Monday, the National Election Data Archive will be releasing its mathematical logic Proof (not just evidence) that the analysis that Mitofsky presented at the Freeman debate on October 14th by presenting scatter plots and analysis by Liddle, plus all the analysis of the Ohio exit poll data by the Election Science Institute since June - which Mitofsky and ESI used to wrongly declare that "this kills (or rules out) the vote fraud argument" - are based on a completely illogical and invalid hypothesis.

Please, if any of you know Ohio press that we could call, or you could please call for us, we'll want to get press coverage of this, so that the public learns that Mitofsky's latest argument has been mathematically PROVEN to be illogical. The democratic leadership seems to have wrongly relied on the Election Science Institute (ESI) analysis too.

Here is our current draft - please keep in mind that it is still considered a draft until Monday when we put out our press release:

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

It proves that Mitofsky/ESI/Liddle have been using illogical, invalid analysis and that their entire analysis is therefore meaningless and should be ignored.

Please sign up to receive Monday's press releases by emailing:

[e-mail] and replying to the confirmation message.

In a few days afterwards, probably by Wednesday, we'll be releasing some VALID analysis of the Ohio exit poll data which clearly shows evidence of vote miscount in Ohio.

We have also updated our history paper to relect the current status of the debate on the validity of the 2004 presidential election.

BTW, this math proof can be understood by and determined to be correct by any math person who understands math logic and arithmetic, so please show it to math professors and the like, to verify its correctness. It requires no background in exit poll science, just a little logic-ability.

Best,

Kathy

[Epiphany4z]

1. I don't know

any one in the press but I have this handy little link...http://www.50states.com/news/ohio.htm

[occuserpens]

2. ESI hypothesis

<The ESI hypothesis, proven herein to be logically incorrect, is stated in an ESI paper as follows. "If systematic fraud or error in vote counting occurred in 2004 but not in
2000, Bush would have done significantly better in those precincts in 2004 , and we would see larger differences between the reported vote and exit poll in those precincts .">

This ESI looks like meaningless propaganda to me, nothing to prove or disprove. What happens is that once we have vote-rigging, crooks can produce any numbers they want.

In fact, this became international standard, the last example is in Iraq. There they concocted the swing Sunni province (like OH and FL) and then cooked the numbers they wanted: 50+% against the constitution instead of 67% necessary by the rules to reject it. In the US, 10-15% is a lot, but in Iraq it is same too close to call. As a result, nobody trusts this abracadabra. The war goes on.

[autorank]

55. ESI said New Mexico was find. Wait until next week to see some

heat on that. There are two cases in New Mexico that show how bad the election was there. ESI trotted down some folks, took nice candid shots and said, "well, it's just fine here." Their noses have been growing ever since.

ESI is a ringer, not all of them but the organization as a whole.

Are there any users here associated with ESI?

[IndyOp]

3. The Free Press & The Cleveland Plain Dealer

Bob Fitrakis & Harvey Wasserman, the "Woodward & Bernstein of the 2004 Election" -- Rev. Jesse Jackson

Bob Fitrakis' Recent Articles:
<http://www.freepress.org/columns/display/3>

Bob Fitrakis & Harvey Wasserman Article:
<http://www.freepress.org/departments/display/19/2005/1534>


Cleveland Plain Dealer has done some decent reporting on Ohio politician scandals and some pretty fair reporting on Election 2004 -
<http://www.plaindealer.com/page2.html>

Best wishes!

:bounce:

[texpatriot2004]

4. Woo Hoo! Kicked and Recommended. Thanks for the link to

the updates.


:woohoo: :woohoo: :woohoo:

[The Doctor.]

5. Tia would love this...

*ahem*... 'Tia Carerra' that is...

(I hear tell she's real good at math)

[texpatriot2004]

6. I heard that too :-) nm

[Melissa G]

7. Sent to the land of tia, my aunt of course, for comment....

[The Doctor.]

9. Great! Be sure to tell us what your

aunt Tia has to say.

Thanks!

:patriot:

[sunshinekathy]

8. I've added a short section on the math proof in the history paper

this history paper on the validity debate on the 2004 presidential election --

http://electionarchive.org/ucvAnalysis/US/Presidential-Election-2004.pdf

was just updated to include a brief section on the proof that ESI/Mitofsky's exit poll analysis since June has been entirely based on an illogical premise.

The section in the history paper (near the bottom of the table of contents) might be easier to read than the actual math logic proof paper itself.

Anyone who could help me call Ohio press on Monday to let them know about this proof and get them to give it to their favorite mathematician to check, would be "really" appreciated.

Thanks.

Kathy

[The Doctor.]

10. I have a new axiom...

"You can't fool numbers."

I'm sure there are some mathematicians who might disagree, but this saying shall apply to statistics... not quantum mechanics.

[Land Shark]

11. Seems to me easy to see that if you increase the size of the pie ESI is BS

If one increases the size of the turnout "pie" in a precinct relative to 2000 (turnout was higher overall) none of the percentages need to change relative to 2000 in order to give the 2000 winner an advantage. (the "bigger pie" effect) that one's a no-brainer for any lawyer on a contingency fee percentage, the case has got to be a certain size in order that the 35% (or whatever) will pay for both the cost of that case as well as any the attorney may lose on contingency and not get paid at all for.... Thus, even though percentages do not change AT ALL RELATIVE TO 2000, in a Bush-leaning precinct, higher turnout means more net votes for Bush, just like a bigger "pie" size of a case means more $$ for the lawyer at 35%.

Example, in a precinct of 1000 voters actually voting, in 2000 Bush gets 600 votes or 60%, Kerry 400 or 40%. Turnout is higher in 2004, with 1200 voters actually voting. The real numbers are 50/50 with Bush getting 600 votes or 50% and Kerry getting 600 votes or 50%. however, fraud "adjusts" these numbers to equal Bush's 2000 percentages, so Bush gets 60% of the 1200 or 720 votes, and Kerry gets the remaining 480. N.B. Fraud has taken place, but Bush's percentage HAS NOT INCREASED from 2000 to 2004, yet Bush has obtained a fraud benefit of 120 votes in his favor, and Kerry has lost 120 votes, for a net fraud benefit of 240.

Now, regarding the above, ESI would say if the exit poll (showing a 50/50 race shows a large Bush red shift because Bush gets 60% instead of 50%, then there should be a correlation with Bush increasing his percentage of the vote between 2000 and 2004. BULLSHIT.

I think in the recent threads regarding "How I stole the 2004 election ..." or another quite recent one, Febble points out something to the effect that 3rd party switches or transfers of votes present interesting counterpoints to E-M because E-M's percentage share numbers assume that only Kerry and Bush exist. (see the exact post, numbered most recently, was in the last 1-2 days) If this means that the Kerry Bush percentages as calculated are zero sum games and don't consider third parties and undervotes, one way to destroy Democratic votes without affecting percentages as much (relative to 2000) is to transfer them to 3rd parties.

I think Febble would also concede that if votes were taken nonrandomly (such as primarily from high Dem precincts) E-M numbers could be consistent with a fraud scenario.

The paper looks fairly good (i'm no math expert) but the tone recommendation on my part would be not to argue it was if you are trying to get out of a tight box. I don't think E-M really PUT YOU into that box in the first place, notwithstanding a statement made at the Freeman - Mitofsky debate and the ESI work. But, this little bit of defensiveness is not really harmful for those tracking this debate.

another possibility. Of course, exit polling is itself open and obvious during the election day. Although the precincts are supposed to be kept secure, if one knew even a decent number of the precincts being exit polled it would be a reasonably simple matter to obfuscate those precincts. However, that obfuscates rather than helps to prove fraud. But if "obfuscation" could be proved, (some hacker or rigger avoided exit polled precincts or otherwise messed with them to obscure other activities) then that would be significant evidence of intent to cover something up.

[Melissa G]

12. Thanks Land shark for this good analogy...I have been trying to get this

point out about the false bush votes not necessarily correlating to any single thing, since they were phantom and not necessarily relative to any single thing, for a while. Yours is the best shot at this I have seen so far.. But as you, or your evil twin :evilgrin: pointed out previously, the theft was all done in so many different uncoordinated ways...

The fraud is not one thing so it does not need to correlate as just one thing even though it does eventually manifest as a net Shrub gain. A bigger pie can hide many thefts, esp with no real valid check on the ceiling of the number of votes or where they might come from.


edit clarity

[OnTheOtherHand]

16. it is worth considering the context of the argument

It's not hard to think of ways that fraud could occur without showing up in the exit polls -- although it may be harder than you think to make them plausible as real-life events.

But we have been told, over and over, that the exit polls are strong evidence of fraud. Yes? that's not just something I imagined?

ESI said that, in Ohio, the exit polls were not strong evidence of fraud. When the "smoking gun" paper comes out, we can see whether ESI has been refuted on that point. But it hasn't happened yet.

The "logic proof" paper unfortunately adds nothing to the debate, except for a cool table. (The table provides a numerical example that if fraud were heavily concentrated in precincts where Bush would otherwise have done much worse, that could reverse the expected relationship between bias and swing. Presumably everyone trained in statistical inference knew that -- it is an application of the basic conceptual framework of why correlations don't prove causation -- and the specifics had already been discussed on DU. But no one had presented a specific example.)

occuserpens is of course correct that if the Bushies have complete control over the vote count everywhere, they can make it turn out however they want. And you are of course correct that the Bushies could have tampered with the exit poll, too. One can't go very far down this path without eliminating the point of "exit poll arguments" altogether. Maybe we are already there.

[BrklynLiberal]

13. kick para mia tia maria...

[sunshinekathy]

14. Some examples

If you read our math logic proof paper, there are some examples in the "English language section". Please read the abstract, the english-language section and the conclusion and then study the two counter-examples.

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

Here are some reasons:

What if the Dems won the 2004 turnout battle big-time so that the effect of vote fraud was to rescue Bush from a worse performance than in 2000 and bring him up to even? Increased voter turnout was 20% in Ohio in 2004.

Or voters may have changed their minds since Bush's first term and voted less for him. In that case vote fraud would bring Bush back up to just beyond even, but certainly there is no reason that vote fraud would mean that Bush had to perform better than in 2000.

The same with Nader voters. Are we assuming that all the Naders voters split evenly to both candidates or did more of them potentially vote for Bush.

Because we don't know what the actual vote was (since we don't routinely independently audit vote counts), vote fraud could have begun at any level and brought the Bush vote up to any level - no reason at all for Bush to beat his prior election shares to have vote fraud.

I've created a spreadsheet that generates counterexamples (but is not in decent shape for public release), but there is a clear table showing one counterexample in the proof paper that uses very similar margins to the actual Ohio 2000 and 2004 election, to show how Kerry could have won (if voter intent were followed) and vote fraud could have changed the outcome, and yet there is no positive correlation with increase in Bush vote share and exit poll discrepancy as Mitofsky, Liddle, ESI's claims that there should be if there is vote fraud.

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

If only I had help and funding (we've been working on our 501-c3 status and a bunch of other stuff - like the history paper on the 2004 presidential election, etc.), I could have gotten around to doing this proof much sooner - but I suppose it's fun to wait until after Mitofsky publicly gave his invalid analysis to the American Statistical Association meeting to show everyone how wrong his analysis is. But really keeping up with the shoddy hypotheses and analysis coming from ESI, Mitofsky and Liddle are wasting a lot of time.

We've got to begin taking action soon though if there is to be enough time to protect the 2006 election.

http://uscountvotes.org/ucvAnalysis/US/election_officials/Audits_Monitoring.pdf

Best,

Kathy

[Febble]

15. Well, first of all

we have a big straw man problem here.

Straw man 1:

You state the ESI "hypothesis" thus:

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}

and, by implication, Mitofsky's "new" hypothesis also.

This statement is indeed false, as you show. However it is not the hypothesis, which I will re-state below (I don't know what you mean by the word "significantly" so I'll omit it):

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

.

This statement is now true.

It is not refuted by your counter-example (1) because of course it is perfectly possible for Bush to do worse in a fraudulent precinct than he did in 2000. However, he won't do as badly as would otherwise have done because, obviously, he benefited from the fraud.

The correlation simply tells you that the discrepancy in the poll did not account for any variance in his vote share over and above the prediction made by his vote share in 2000. It's what correlational analysis is all about.

Straw man 2:

You repeatedly allege that Mitofsky, ESI, Liddle (i.e. me) and now Lindeman have said that the analyses "rule out fraud". Well, I'm not going to speak for anyone except me, but DUers at least will know that I have said no such thing. Nor, to my knowledge has Mitofsky. What I believe Mitofsky said (and what I happen to know he meant) was that the plot "kills the argument that fraud was the cause of the exit poll discrepancy". It certainly doesn't rule out fraud - many types of fraud will not show up in the exit polls. Many of those types of fraud occurred, I believe, in Ohio, although some don't always go by the name of fraud. But there are also types of actual, full-blown, electronic fraud that are not ruled out by the analyses - but thye would only account for Bush's win, not for the exit poll discrepancy.

Straw man 3:

Me. I'd appreciate it, Kathy, if you'd read my recent threads (also quote them accurately in the paper, preferably with links - use copy and paste so you don't misquote) you'd see that I have been actively searching for ways in which the analysis you dismiss, rightly, as failing to "prove there was no fraud" could be compatible with certain types of fraud. I agree that if fraud was concentrated in large precincts in which Bush was doing badly this could escape detection in the correlation and yet garner precious votes for Bush. However, it would still need to be limited numerically, as ideally it should be concentrated in precincts where not only Bush was doing badly, but also the exit polls were going to be a bit blue-shifted (there were plenty of those) and although you might be able to predict the first, you'd have a job predicting the second. If you only target precincts where you think Bush is going to do badly, you will still induce a correlation unless you target very few.

I've been doing some modelling, and discovered it's pretty difficult to avoid producing a correlation between fraud and vote-swing even if you concentrate the fraud in Bush's disaster areas. Certainly large precincts would help. This suggests a testable hypothesis: is the correlation between swing and PLD greater in large precincts than small?

So there goes straw man number three - I'm not arguing against fraud - I'm arguing that it is not indexed by the exit poll discrepancy (not "indicated" as you quote me in your footnote).

So enough already with the confrontational tone, Kathy - please write a paper that outlines the kinds of fraud compatible with the ESI and Mitofsky findings, which we could all use. LandShark's thread, and mine have some good ideas (see the comments by people like TfC, BillBored, eomer, anaxarchos). But please stop haranguing people for making points that they didn't make.

It might also be worth checking out the points they DID make, as my contention is that these analyses actually HELP us look for the right kind of fraud in the right kind of precincts (or the right kind of algorithm).

[Febble]

17. Correction: "let me read that again"

Hypothesis is:

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

.


In other words: Bush, in the absence of fraud, is likely to do better in some precincts relative to 2000 and worse in others. In the presence of fraud, he will do better than he would have done without fraud. Obviously. But this could as easily been in precincts where he was doing terribly (so it would just have made his performance less bad than it would have been) as in precincts where he was doing well (in which case it would have made a good showing excellent.

Kathy is quite right to point out that if fraud was concentrated in precincts where he was likely to do badly (and BillBored has made the good suggestion that precincts with big increases in Dem registrations might have flagged up to fraudsters good precincts for fraud) then the correlation between Bush's performance relative to 2000 and the PLD might be at least partially masked. And it is possible that with the relatively low signal to noise ratio in the ESI study that quite a lot of fraud could have got under the statistical radar if it had been done this way (although there are other checks - for example the relatively good match between polled precinct counts and statewide precinct counts). I do think this is worth exploring.

However, the signal to noise ratio in the whole sample is good. If we look at Mitofsky's second plot:




we can see that, on face value at least, it appears to suggest that greater redshift was not associated with greater benefit to Bush. The regression line is absolutely flat. But, as DUers will know (though Kathy may have missed some of this discussion) some of us (including me) have been trying to figure out how fraud might account for Bush's win (and the red shift possibly) without producing a positive correlation between swing and redshift. And I have to say, pace Kathy, that it's not a slam dunk.

Let us consider the plot: each data point represents one pecinct. It can be considered to have four quadrants. We have two quadrants in which Bush does better than his average gain relative to 200(above the zero line), one where the PLD was redshifted, and a slightly smaller one where it was blue shifted. Similarly we have two quadrants in which he does worse, the larger one of which is redshifted and the smaller blueshifted. Kathy, I think, is proposing that fraud was concentrated in precincts below zero (the two lower quadrants), and resulted in both redshift (as it must) and a rather better vote-share. for Bush than he would otherwise have got.

So, in effect, we are saying that the "true" position" of at least a proportion of precincts in the lower two quadrants is further down and further left. "Fraud" has shifted them upwards and rightwards into their current positions. If we shift them back into their "true" positions, mean PLD will go back towards zero, and Bush will do less well.

The problem here, which is what I have been wrestling with for the past week or so, is that if the "true" position of those precincts is as I have described, the regression line would actually shift positive (we'd be moving a bolus of precincts from their current position below the line down and left, pulling the left hand end of the regression line down with it. In other words, for this hypothesis to work, we'd have to postulate something rather odd about the exit polls: that had fraud not occurred, precincts in which Bush did relatively well would tend to be "redshifted" and precincts in which Bush relatively badly would tend to be "blueshifted".

So, we'd have to postulate that Bush voters were more reluctant where Bush had more enthusiastic support, and Kerry voters were more reluctant where Kerry voters had more support. Which is a bit of an odd postulate, though perfectly possible, I suppose.

But we certainly don't have any evidence for it, other than it would allow fraud to account for the mean redshift.

But certainly if fraud was targetted at relatively few precincts, preferably large ones, it would have helped Bush and maybe not impacted on the plot.

But in that case fraud wouldn't account for the redshift. Which is what I understand Mitofsky was saying.

On Edit:

It has just been pointed out to me (by OTOH)that if we assume that fraud only occurred in the bottom right quadrant - that all other PLD, including that net redshift in the top right quadrant - was polling error, then the only "shifted" precincts are now in the bottom right. If we move those down and left to their "true" positions, we probably won't shift the regression line - if anything we will tend to shift it positive.

OK, we can work with this.

[Land Shark]

18. could you address the pos and neg effects of increased turnout

on your analysis?

You allow as to how targeting Hi Dem precincts could help mask the correlation, but suggest that hiding enough of it may be problematic.

Does higher turnout have a masking effect under any circumstances?

Does shifting votes to third parties and/or to undervotes have a masking effect OR other interesting properties? (Note the obvious: a direct vote shift is -1 Kerry and +1 Bush, for a net benefit to Bush of 2, while a shift to undervote is -1 Kerry and 0 Bush, for a net benefit to Bush of 1, meaning that the effect AS BETWEEN Kerry and Bush is masked by 50% in terms of Bush GAIN, BUT THE REDSHIFT SEEMS TO REMAIN THE SAME BECAUSE ONE KERRY VOTER HAS "DISAPPEAREd" FROM RESULTS COLUMNS IN THE EXIT POLLS)

Thus, it seems, does it not, that if E-M is using only Kerry/Bush percentages adding to 100%, this distorts all their figures, and causes nonrecognition of the 50% masking effect of switched votes to third parties or undervotes.

[Febble]

19. Hoo boy

this is getting complicated. Certainly, shifting votes from both Kerry and Bush to third parties in high Dem precincts would be is a good trick. It wouldn't mess up the WPE, but I am trying to figure out what it would mess up.

And turnout - I'm thinkin', I'm thinkin'....

(being a criminal is HARD WORK....)

I've just convinced myself that if we postulate that the extra redshift in the bottom right quadrant was due to fraud, it might not produce a linear correlation (but it produces a visibly odd distribution at present, and but maybe I can finesse that). However it leaves the redshift in the top right quadrant not accounted for, and that doesn't make a lot of sense. Why would you get polling bias only in precincts where Bush was doing well?

However, if we account for the extra redshift by fraud (by squeezing the red dots towards the zero bias whether or not they lie above or below the zero swing line, then we get our correlation back. So I'm still left with the conundrum that if we want fraud to account for the extra redshift in a coherent manner, we get a correlation, but if we don't we don't. But then we don't have exit poll evidence for fraud (which is fine by me, but probably not by Kathy).

And using intuition here, I'm still back where I was a few days ago - any one thing, on a big scale, is going to show up somewhere. So the best hypothesis for Grand Election Theft is salami slices here and there, always keeping within the statistical noise.

Back soon.....

[Land Shark]

20. That's why they call it "vote shaving"....

EVERY WHISKER DROPPED IS NOT ENOUGH 'TO CHANGE THE RESULT' BUT THE END PRODUCT IS A SMOOTH, CLEAN SHAVE AND A SLAP TO DEMOCRACY. mmmmmmmm.... THANKS, I NEEDED THAT! (warning: television allusion to old commercial)

That being said, not every technique is just the equivalent of a single whisker, the undervotes method is going to produce a fair bit of stubble on its own.

[Melissa G]

23. LOL!!! that's a blast from the past!

Just watched the original Philadelphia Story last night..There was some dialogue!.. my younger daughter raised on today's excuse for movies and TV ambles from the room.. not used to needing to pay attention to follow what's happening...

ah, how far we have fallen as an artistic culture...how far we have fallen in so many ways...

[Joe Chi Minh]

49. Your remarks remind me of the reaction of the mother of

a student here (which she described in a letter to the Times (UK)), on learning that her sprog had just graduated: "Thank God for falling standards..."

[Febble]

71. Yes. Nice.

And in fact it is perfectly possible that the entire election was stolen without impacting on the exit polls at all.

Which just happened to be inconveniently wrong for their own sweet reasons. Or else they were planned as a deliberate misdirection from where the actual business was going on.

[Land Shark]

73. What matters in the end is whether the sheep have been shorn

[Melissa G]

74. Nekid Sheep!!! Nekid Sheep Everywhere!!! n/t

[Land Shark]

75. Baaaaaaad sheeple jokes are always in order! nt

[Melissa G]

77. Very BAAADD!! ROTFLMAO!!

:toast: :rofl::rofl::rofl::yourock:


Update for another smilie...

[Land Shark]

81. Melissa, Now there's all KINDS of dirt ground into your short coat....

and there's enough shavings to pull the wool over the eyes of every other American.

LandShark's evil twin LOVES this scenario's elegance: If someone could in fact also obfuscate the exit polls, it would be the perfect crime, because in a land where people are normally falling all over themselves to apologize for even *suggesting* that anyone under cover of secrecy would agree with another to try to accomplish an illegal goal (the legal def'n of conspiracy) is UNTHINKABLE. Even though it happens thousands of times a day. And surely, there are no precedents in world history for any shenanigans done in the name of world control, are there????????

Because it's unthinkable for so many, even a smoking gun might not convince them, so long as superficially plausible talking points existed against the smoking gun. It's just not kosher to think that people work together on illegal aims. Though as a matter of fact, conspiracy is one of the prosecutor's most potent tools, and not all that hard to prove either.

Admittedly, talking about a possible conspiracy is basically a thought crime in the "land of the free and the home of the brave". Therefore, it's endlessly popular to bash conspiracy theorists for having the gall to engage in educated guesses about what has been deliberately made secret by others.

At this point, the "Election CSI" folks if they existed would definitely consider the exit poll data itself a suspect in the case. But then that would annoy BOTH Febble/OTOH as well as USCV, (for different reasons of course).

Yet, with this level of unexplained discrepancy, and with the discrepancies ALONE being cause to suspect BOTH the election and the exit poll, I think Febble is compelled to admit the inference of exit poll tampering is WIDE open. (I'm not saying "proved"). I know she doesn't want to "go there". Still, I think that Febble at least, when intellectually honest, would have to agree:

WE CAN NOT *RULE OUT* TAMPERING WITH THE EXIT POLL DATA. Febble says she can't "imagine" creating such a "real" feeling data set, but when we have "real" data assumed from 2000, why this is such a leap of imagination escapes me.

Where does this leave us? Staring at the abyss, nothing but a black hole with a some sheep shavings floating around aimlessly disconnected. When the free people of the USA let their elections out of their sight and supervision, the tether to the mother spaceship of democracy was cut and we drifted free into the void. We're all just standin' around Nekkid, Nekkid, Nekkid as sheeps. Then the guys who cut the tethers of democracy tell us we got little or no EVIDENCE of who the hell did it or why. So, most of us if we even see the cut tethers dangling down from our 1 hour spacesuits, say "rats, foiled again" and settle down to die of exposure in deep space, having had our choice made for us from the Revolutionary motto of "Live Free or Die".

[Melissa G]

91. No wool left on me LS, I'm a Nekid little sheep...Like democracy

on life support floating in space..Nekid..

We have massive distortion in exit polls...no smoke there..
We have this yet to be accounted for gender shift..nope nothing there..
Optimizers run in a galaxy far far away from DU..still nothing...

There is still no plausible theory that accounts for the the strange anomalies other than Fraud.. but nothing there either... there no story line to follow just a theory of reluctant voters that is just that..a theory which someone likened to believing the theory of intelligent design as science.. I can buy that.. but it won't keep me warm cuz I'm Nekid...

BAAA it's cold out here in deep space..Nekid...

[Melissa G]

92. Democracy has done been fleeced at least three times now..

The fleecing is now even institutionalized into the way we interpret our ever evolving exit polls.. We know there was likely Fraud in 2000 and in 2002, Yet these elections have become the baselines for our discussions.. So we Nekid sheep are saying we are not shorn... let's compare to 2000!.. Let's compare to the fleecing of 2002! and now we will use the Nekidness baseline of 2004 to further convince ourselves that we are not cold and dying in deep (doo doo) I mean space...when it comes time to evaluate the 2006 and 2008 elections ... Yep, pretty much F****ed, I mean Fleeced..

edit clarity

[Land Shark]

122. I wish the people who fancy themselves our leaders were understood

as typically NOT wanting a fair and democratic process, because this is in fact what happens 98% of the time in corporations, organizations, even nonprofits. EVERY ATTEMPT is made to play the vote like a game, instead of informing the public or the workers as much as possible and then respecting the decision of the whole. That latter possibility seems fairly Fanciful to imagine as our reality, yet it is the nearly unquestioned myth of the biggest elections of all: for President.

In reality and as a general rule, the "threat" of democracy is responded to by leaders of all kinds including political leaders professedly in support of democracy with every conceivable attempt to lobby, cajole, distort, or otherwise prevent the vote of the "people" of that organization from being any different than what the leadership desires.

Why would all this pressure (legitimate and illegitimate) not be brought to bear on the biggest elections of all?

And putting blinders on the people is nothing new. Nothing new at all. It's just another and very effective way to do this, to institute electronic vote counting.

[Land Shark]

21. the problem with using percentages is also that Bush's percentage

could drop in a given precinct, but greater turnout pie could mean more of a net benefit to Bush in terms of the total lead he emerges from that precinct with.

precinct X, 1000 voters in 2000
Bush share 60% nets him 600 votes, relative to Kerry's 400, net benefit to bush is 200 vote lead

Precinct X, 1200 voters in 2004
Bush share of 55% nets him 660 votes, a gain of 60 over 2000
while Kerry's share of 45% gets him 540 a net gain over his 2000 performance of 140 votes. Bush has dropped 5 full percentage points but Kerry picks up only 80 votes relative to the bottom line of 200 vote advantage for Bush in 2000.

Same Precinct X, again in 2004, this time Bush gets 58%, down only 2% from 2000 instead of 5% as in the example immediately above.
Bush share of 58% nets him 696 votes in this precinct, while Kerry's share is 504 votes. Now the bottom line is a 192 vote advantage for Bush over kerry, almost identical to 2000 even though Bush has lost 2 percentage points to Kerry.

I know Febble and OTOH and SunshineKathy and any other respectable mathematicians can devise a formula, the variables for which are "INCREASE TURNOUT FOR PRECINCT" in question, BUSH2000%, KERRY2000%, BUSH2004% and KERRY2004%. then, for a given increase in turnout based on known 2000 performances of Kerry and Bush, you can calculate ON A PRECINCT BASIS (Because iT'S DIFFERENT IN EACH ONE) where the break points are in terms of creating or not creating red shift correlation with gain.

Now, if LShark's evil twin combines these precinct breakpoints with detailed polling data (assume it's accurate of actual turnout) OR in the alternative combines these precinct breakouts with actual real time results at some point preferably during election day or at the tabulator level, then one can divine rather quickly, with the aid of computers, precisely which precincts to tinker with and by how much to avoid smoking guns in the traditionally expected sense, yet still add/hide a lot of net benefit to *.

[Name removed]

22. Deleted message

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

[sunshinekathy]

35. This posting is incorrect

The quote comes exactly from ESI's paper, so claiming that you can remove words from it, and that it is incorrect - well go read ESI's paper and stop misrepresenting it.

The hypothesis as you restated it would be impossible to evaluate or do any analysis based on because we don't know exactly how the voters voted - so obviously that is not the hypothesis that you based your analysis on.

If you read ESI's paper or attended the Freeman-Mitofsky debate, both claimed that their illogical invalid analysis "ruled out vote fraud" so to claim that they didn't use it this way, is not the truth.

Why would anyone waste time doing analysis that "nothing" can be concluded from because it is based on an incorrect illogical hypothesis, is more than I could fathom - unless there is an agenda.

Please tell the facts the way they are.

[Febble]

44. Please read my response to Sean

It should have been posted to this comment.

But the hypothesis I stated is perfectly possible to test, and indeed it was tested by both ESI (however they mistated it). It is the essential statement of a regression analysis:

If A and B vary and C causes variance in A and B, A will be correlated with B.

Variance in A that is not caused by C will be unshared with B
Variance in A that is caused by C will be shared with B.

The proportion of variance in A and B that is shared, will tell you how much of variance in A is caused by C.

If A is PLD, and B is swing and C is fraud, the proportion of variance in A shared with B will tell you the amount of variance in A that is due to fraud.

This hypothesis rests on certain assumptions which have been challenged by many, including myself. But the fact that A and B share no appreciable variance means that if fraud occurred, we can deduce something about the violation of those assumptions that must have occurred.

[Febble]

70. The ESI hypothesis is correct

without the addition of your bracketed interpolations. I was mistaken in my posts downthread that they had misstated the hypothesis - I failed to spot that your interpolation does not appear in their text.

You rightly falsify your edited version of the hypothesis.

However, the hypothesis as stated in the ESI report is perfectly logical. And quite a lot can be concluded from the fact that the statistical test used indicated that it was not supported.

Your implication that anyone testing such a hypothesis must have an "agenda" is offensive.

I accept that you think that you have proven that the hypothesis is illogical. I accept that you think that the reports were "hurried" and "irresponsible".

I submit that it is possible that your posting of a draft refutation may also prove to be "hurried", not to say "irresponsible" - and that your paper does not, in fact, refute the logic of the hypotheses tested.

However, what I believe IS irresponsible is publicly posting a paper that mistates other researchers' hypotheses and claims to demonstrate that those hypotheses are not simply "illogical" (which remains possible, and a matter that we might debate in a civil fashion) but that those who test those hypotheses are in some sense complicit in the covering up of fraud.

And as I am specifically named as one of those researchers I will state here and now that I reject that accusation.

I am happy to be proven wrong (although I am as yet quite unconvinced that your "proof" that I am is any such thing). I am not happy to be continually maligned as a producer of "shoddy" "irresponsible" "hurried" "misleading" work and having an "agenda".

Enough. Now.



On edit: I accept that the paper itself does not claim to demonstrate that researchers testing the hypothesis are complicit in fraud. It is this post implies so and it this implication that I reject.

[sean in iowa]

38. Here is a link to the ESI report and the quote discussed here:

See page 3, the top of the right column.

Here also is the quote, along with an important sentence:

Finally, ESI examined whether the proportions of the vote that Bush received in each precinct, in 2000 and 2004, were related to the difference between the reported vote and exit poll results for those same precincts. If sytematic fraud or error in vote counting occurred in 2004 but not in 2000, Bush would have done significantly better in those precincts in 2004, and we would see larger differences between the reported vote and the exit poll in those precincts.

Febble's explanation of the ESI hypothesis does not make sense to me either.

[Febble]

41. Fair enough

In that case I disown it. Kathy presented it as:

"The ESI/Mitofsky/Liddle conclusion that the nationwide exit poll data shows no evidence of vote fraud in the 2004 presidential election4, is based on the inference that...." etc.

It is certainly not the inference I based my conclusion on.

My hypothesis was as I stated.

Moreover, my statement of the hypothesis serves as a perfectly good statement of the ESI hypothesis (they seem to have messed up).

[Febble]

24. It has been brought to my attention

that I should make it (once again) clear that an analysis is one thing; interpretation is another.

The analyses presented by Mitofsky and ESI indicate that there is no linear relationship between the degree to which Bush improved/failed to improve his vote-share in 2004 relative to 2000 and the degree to which the PLD (Precinct Level Discrepancy in the exit polls) departed from zero. These are very important findings, and a simple reading (even quite a complex reading) suggests that the fraud made little or no contribution to the exit poll discrepancy. This is not the same thing as saying that there was no fraud. It is perfectly possible that the exit polls were "wrong" in the "right" direction for quite coincidental reasons.

But the reason the analyses are important is that they do indicate that IF fraud occurred in the 2004 election, it must have been either of a certain type or targetted in a particular way. And this should help us look for the most likely kinds of fraud.

As far as this goes, I have no problems with Kathy's paper, except that she repeatedly mis-states my own interpretation, and I am, I confess, getting a little fed up with the role of bad guy in all this.

So to be absolutely clear: Kathy quotes me, in footnote 4 as saying:

"the failure to find a linear relationship between the magnitude of the exit poll discrepancy and the magnitude of Bush's increase in vote share since 2000 is a major problem for the argument that the discrepancy indicates{sic} fraud."

I presume the origin for her citation is this:

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

and I stand by this statement, with the correction that I used the word "indexes" not "indicates": a small difference but an important one. I will therefore paraphrase my statement thus:

"The failure to find a linear relationship between the magnitude of the exit poll discrepancy and the magnitude of Bush's increase in vote share since 2000 presents a major problem for the argument that the discrepancy is a measure of the magnitude of fraud"

And will add to it: It does not, however, rule out fraud, although it sets important and informative constraints on the type and target of such fraud .

It's these constraints we have been exploring recently on DU, and would certainly be worth writing a paper about.

[Melissa G]

25. Personal Thanks To EVERYONE who is working on exposing Fraud

in the 2004 sElection of the current occupant in the White House.
You are all great Patriots whichever country you live in including England and New Zealand...
We all owe everyone a debt of gratitude for their efforts. Let's spend our energy constructively finding the causes of possible fraud and working for the solutions which insure the next election will be fair, transparent and verifiable rather than wasting our time in less fruitful personal attacks. That said, Ideas (not people) that are in need of serious revision are fair game..

[sunshinekathy]

29. Some of us are sticking to the truth & good logic though

Some are putting out invalid analysis without checking their hypotheses, either logically, algebraically, or analytically before publicly releasing it (like Mitofsky, ESI, and Liddle) and some (like NEDA) are being more responsible.

I would hesitate to thank those who are throwing out hastily done analysis without checking it first, and therefore misleading the American public into thinking that there is no evidence of vote fraud in the 2004 presidential election results.

[Melissa G]

36. ESI is a group not a person and I am not thanking them. They are quite

suspect IMHO. E/M is a group also not a person that I am not thanking. Warren does not post here and neither does Mark Blumenthal and I don't see them working on exposing fraud anyway so no comment needed..
Febble is a person who posts here and I mostly disagree with her. She does however, in her own way, work on exposing fraud. She is also courteous and thorough. If you think she makes a mistake she will check it out. She also checks out other people's work. A different point of view is useful in a discussion, so for purposes of furthering discussion, I'm glad she and her polite ways are here. I will remind you that DU rules and civilized discourse also require it.

If you check my posts you will know that I am one of the first people to jump on something that I consider misleading....or rude.

[Name removed]

61. Deleted message

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

[Febble]

62. Kathy, your proof

does no such thing.

Now: I have finally found my proper specs, and read the ESI statement, which although not ideally formulated is actually fine without your interpolated phrase: "than in 2000".

I appreciate that you have not said I am a con artist, and I agree that "bandying about a few good-sounding math words" does not mean that a person is doing a decent job.

But "bandying" is in the eye of the beholder, and I simply do NOT consider that your "Proof" is any such thing. In fact the word "bandying" rather springs to mind. The hypothesis as stated by ESI is NOT falsified by your counter example 1 for the simple reason that the hypothesis does NOT (in ESI's words or mine) say that fraud will lead to Bush doing better than he did in 2000. It simply says that he will do "better". Which obviously he would. There is no point in fraud unless Bush does better.

To which you respond that this is not testable. Well it is. It is testable by exactly the test ESI and Mitofsky used.

Now those tests rest on certain assumptions which may be violated - in particular they assume no interaction between fraud and swing - and there may indeed by such an interaction if the fraudsters were clever enough to target fraud in precincts where, say Bush was doing badly.

But these blanket assertions that you have "proof" that the actual "hypothesis" is "illogical" are, IMO, not supported. It is an absolutetly standard statistical hypothesis-with-test. And to accompany your assertions with aspersions on my "diligence", and implications that those who trust that I am trying to do good work have been "fooled" (because Melissa does not actually say she has agrees with my conclusions, merely that she considers me a genunine fraud investigator), is to say the least, offensive.


On edit: BTW of course Mitofsky is listed as an ESI "author", although you will note that he and Scott Dingman are listed after the word "with". They prepared the "blurred" data. It is usual in a study to list those who collected and prepared the data as authors. In this instance, to clarify the distinct roles, the data collection/preparation team are relegated to "with". Obviously ESI couldn't have done the study without Mitofsky and Dingman, and if NEDA ever succeeds in getting the exit poll data released for analysis, you too will be listing, as an author, Mitofsky and whoever he is currently working as his statistician (not me BTW), presumably after the word "with".

(also edited to supply missing phrase in main post)

[Melissa G]

63. Kathy, Where exactly do you get off saying I have been fooled...

Cite Please.
I am no Fan of Mitofsky or ESI.

Read my posts. I am no fan of FFF ( Febble's Fancy Function) as a defense of the exit polls..I am no Fan of FFFV ( Febble's Fancy Function Variant). I have actually told Lizzie that I think Mitofsky uses her skirts to hide in... We manage to agree to disagree about this. She knows how I feel and I know how she feels.
We are civil about this discussion and so there are other things in the world that we can talk about because of this. I respect Lizzie even if I disagree with her.

I'm not yet sure what your proof proves. I posted it on another board where it is currently being reviewed. I am sure it is likely in essence correct. I suspect, like Landshark, that a poor tone and poor word choice may give some folks fodder to shoot your proof down unnecessarily. I can say that as a person who spends a lot of time on this forum while I appreciate your efforts, I don't care much for your lack of civility and esp not your unfounded assertions about me.

If your out of nowhere assessment about my being fooled is any indication of your research ability... then i would say that yes, there may be some problems to uncover..I hope I am wrong about this.

[Land Shark]

26. and I would add, to Febble's clarification and expansion of ideas that

at the end of Febble's clause "it sets important and informative constraints on the type and target of such fraud" it should be added "provided we assume that the nonfully disclosed exit poll data has integrity, and that there was no ability of any potential fraudster to know the precincts that were being exit polled, or to have access to the exit poll data prior to release" (or something to this general effect.

And i agree with Melissa that it is the ideas that are important here, not the personalities.

[Febble]

27. Yes, I would agree with LS's

addendum.

And with Melissa, of course

[Bill Bored]

82. I think it's a veritable CULT of personality myself! :) nt

[FogerRox]

89. >wink< Is that an AgEnDa DA ! ! !

[sunshinekathy]

31. ESI/Mitofsky/Liddle analysis is Meaningless - NOT informative

This math logic proof demonstrates very clearly that the type of analysis being done by ESI/Mitofsky/Liddle, is MEANINGLESS because it is based on invalid illogical hypothesis.

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

NOTHING informative can be learned from doing invalid illogical analysis!

Please read the proof - but I am updating it today to make it more understandable (I hope).

I'm not sure why people do not yet understand this, so I'll state it again:

NOTHING informative can be learned by comparing the 2000 and 2004 Bush vote shares versus the exit poll discrepancies - because, as my proof shows, the entire analysis is based on an illogical premise.

Please read the proof which shows what data would be needed to do any valid comparison of the 2000 and 2004 elections.

[Land Shark]

33. Like I said, I'll read the paper again, but sunshinekathy's conclusions

while clear, are not as useful as understanding the analysis and the data (kinda like elections, eh?). so, if the explanation isn't clear, I can't accept the clarity of the conclusion, no matter how clearly and forcefully stated.

[Febble]

37. Kathy, I've read the proof

and I'm here to debate it with you. Repeatedly asserting that it is proof that what I have done is illogical is not really advancing our understanding.

You appear to have assumed the hypothesis was:

IF systematic fraud or error in vote counting occurred 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 count and exit poll results in those precincts.

And I agree with you that this hypothesis is not logical.

However, I would argue that this hypothesis IS logical:

IF systematic fraud or error in vote counting occurred in 2004 but
not in 2000, THEN Bush would have done significantly better relative to his vote in 2000 than he would otherwise have done, AND we would see larger differences between the reported vote count and exit poll results in those precincts.

In particular, it is not invalidated by your counter-example 1, IMO.

You write:

"if we give a counterexample that shows how Bush can win with pro-Bush vote count error in 2004, and win in 2000 without vote count error, and not do significantly better in those precincts in 2004 than in
2000, then we have shown that A B ® is not a valid inference."

Leaving aside the word "significantly", which I don't understand in this context, it will be perfectly possible for fraud to result in Bush doing worse in 2004 than in 2000, yet not as badly as he would have done without fraud. The point remains that if fraud was accounting for additional variance in Bush's performance over and above whatever natural variance there was, then that variance would be reflected in variance in PLD. It would, in fact, be quantified by the percentage of shared variance between PLD, and swing, which turns out to be absurdly small.

However, I agree that you raise some good and important issues in your next point, which is that if the fraud was specifically targetted at precincts in which Bush was expected to do badly, this might mask the correlation. This is an idea I have been actively exploring on DU with the help of a number of other posters.

So what would be REALLY nice is if we could work collaboratively rather than confrontationally on the issue of the kinds of fraud and the kinds of targetted precincts that would be consistent with the ESI and Mitofsky plots, as a number of us have been doing on various threads recently.

[Febble]

46. OK, truce here:

I agree that ESI's statement of their operational hypothesis is actually wrong. And I do see that you quoted it directly from the ESI paper (although you attribute it to Mitofsky and Liddle as well.)

It is, however, easy to infer what hypothesis is being tested from a given statistical finding, so it is actually perfectly possible to infer what ESI actually tested, which is the one I stated.

So I'd like to know whether you consider that hypothesis to be illogical, as it is the hypothesis that I believe was tested in both analyses.

[sunshinekathy]

28. Those findings are meaningless - not "important"

Clearly, if you understood what the proof showed, you would understand how meaningless those findings are because clearly there is NO inference that can be made from any finding whether there is ANY relationship between Bush's performance in 2000 and 2004 and WPD (exit poll discrepancies).

Your findings indicate NOTHING, because the hypothesis you base your entire analysis on is invalid.

As I showed in the paper, you can have fraud, and yet have a Negative or positive linear trend when you compare WPD with Bush increase in official vote share.

Please get this - Your findings are MEANINGLESS because the entire analysis is based on an illogical premise.

[Febble]

30. Well, I do dispute that, Kathy

which is why I posted. It is of course possible I have misunderstood you, but by the same token it is also possible that you have misunderstood me.

I do not believe that my hypothesis is invalid, as stated as I have stated it, although I certainly agree that the hypothesis you refute is indeed invalid.

I just disagree that that was the hypothesis being tested!

But could you perhaps bear in mind Melissa's point that we are discussing ideas here, not personalities? I'm very happy to debate your logic, but the tone of your post is not very conducive to rational debate.

[Land Shark]

32. I'll have to read your paper again Kathy, but it seems to me that

the findings are based on a premise that is not as broadly applicable as it would need to be to be final, but still has some more limited utility.

If you are making the broader claim, I'll have to read your paper again, which is not easy reading for me.

I personally find, at least as for us non-mathematicians, that a conclusion that *completely* lists all of its assumptions next to a reasonably detailed conclusion is less likely to be misused in the political debate, either unintentionally (lack of understanding of math) or intentionally (for political gain)

Assuming the case is not 100% clear of 100% invalidity of Febble, then it would be a tactical mistake as well do deny the limited utility as defined, because IN THE END the denial of that limited truth will BUILD INTO THE DEBATE a nice hole for experts NEW to the debate (if it should, say, reach the MSM) to hang a quick-denial hat on. That seems to me to be a tactical error.

My bottom line position is I find myself thinking febble still has some more identification of assumptions to do that will further limit the applicability of the redshift/gain correlation argument even though her mathematical intuition was partially correct and valid, and SunshineKathy will not be able to maintain the position of complete invalidity of the ESI hypothesis although her mathematical intuition, too, was correct.

A principle can be inapplicable in 100 places, but still applicable in one or five other places.

[sunshinekathy]

34. There is NO utility from doing INVALID illogical analysis

It is very clear that NOTHING can be inferred from the analysis that ESI/Mitofsky/Liddle are doing which is why I took the trouble to write up a proof of the 100% lack of logic in their hypothesis on which they base all of their recent analyses since June.

It is really very simple - any analysis based on an illogical hypothesis can show you ZIP.

All ESI/Mitofsky/Liddle's analysis has done is create a huge waste of time and misdirected attention and misinformed.

I have PROVEN this with a math logic proof.

[Febble]

39. Kathy, if you are so sure

you have proven it, why don't you engage my point? I am arguing that what you have successfully done is to prove a point illogical that was not actually made.

If you want to continue to assert on all these threads that the hypothesis behind the ESI and MITOFSKY analyses is illogical, it would be helpful if you could address the matter of what hypothesis is actually being challenged.

I am not convinced that you have actually demonstrated the illogicality of the actual hypothesis, merely the illogicality of the hypothesis you THOUGHT was being made.

[sunshinekathy]

40. Read my earlier response to your "restatement" of ESI's hypothesis

And study my proof

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

so that you will understand it.

[Febble]

42. And read my response to your

response.

And I have studied your proof.

[sean in iowa]

43. In this post, you just paraphrased Kathy's supposedly invalid

intepretation of the ESI hypothesis.

You state in this post:

The analyses presented by Mitofsky and ESI indicate that there is no linear relationship between the degree to which Bush improved/failed to improve his vote-share in 2004 relative to 2000 and the degree to which the PLD (Precinct Level Discrepancy in the exit polls) departed from zero. These are very important findings, and a simple reading (even quite a complex reading) suggests that the fraud made little or no contribution to the exit poll discrepancy.

So you are saying that the more Bush gains on his 2000 percentages, the greater the exit discrepancy should be. Yet in above comments you take Kathy to task for misintepreting the ESI hypothesis to be just this.

To quote:

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

.

You've just made two quite different representations of the ESI hypothesis.

[Febble]

45. Well it's a subtle point

which is that the ESI statement (presumably carelessly) implies that if fraud was responsible for the exit poll discrepancies in Ohio, then Bush should have done better in those precincts than he did in 2000.

This is a poor statement of their statistical hypothesis, and it is fair enough for Kathy to have demonstrated that it is illogical. It is. But because it looked as though it was attributed to me, I failed to notice that it was actually ESI's wording. Slap on the wrist for ESI and apologies to Kathy.

But what they actually tested, statistically, was what I stated. i.e. not that Bush would do better than he did in 2000 if there was fraud, but that he would do better than he would otherwise have done if there was fraud.

And it was tested by an Analysis of Variance, also called the General Linear Model. What they found, and what Mitofsky found, was that there was no significant shared variance between swing and redshift.

[sean in iowa]

47. ESI misstated their own hypothesis? Where do they state it

correctly? Could you provide a link?

This is the ESI hypothesis as I understand it. Please respond with a simple yes or no:

If fraud is responsible for the exit poll discrepancy, then in precincts where, in 2004, Bush improved the most on his share of the 2000 vote, the exit poll discrepcancy should be the greatest. In precinct where he his share of the vote declined vs. 2000, the redshift should be lowest, or even of negative value.

[Febble]

50. Sure

page 3

http://www.electionscience.org/Members/stevenhertzberg/report.2005-07-19.7420722886/report_contents_file/

Regarding your question: sort of.

It is actually quite tricky to state a statistical hypothesis in words, although it is actually stated in the terms of the analysis.

You wrote:

"If fraud is responsible for the exit poll discrepancy, then in precincts where, in 2004, Bush improved the most on his share of the 2000 vote, the exit poll discrepcancy should be the greatest. In precinct where he his share of the vote declined vs. 2000, the redshift should be lowest, or even of negative value."

Which is more or less right.

But the point about Analysis of Variance is that you cannot disprove it by a single counter-example. There will be other factors affecting both shift and discrepancy, and sometimes Bush will do badly because he is deeply unpopular - but fraud will rescue him from doing quite so badly. In others Bush will do well for other reasons - whether there is fraud or not.

This is why we partition the variance. The idea is to determine the proportion of variance in Swing that is shared with Redshift. If a lot is shared, it will look as though there is a common factor - fraud. If most of the variance is unshared, then it looks as though there is not a common factor. The statistical test (one of them) is the F ratio - between the variance shared and the variance unshared. In both analyses, the variance unshared was far greater than the variance shared, rendering the correlation between the two "insignificant". In the presence of an "insignifant" correlation we cannot say whether there was a common factor or not.

In the ESI study the number of precincts was small - so the upper confidence limit on the correlation might allow for a bit of a relationship - there might be a "true" relationship of some size, that might be due to fraud.

However in Mitofsky's analysis the full 1250 precincts were included, making the confidence limits on the correlation pretty tight. It really looks as though there was very little shared variance.

Thus far, I maintain both studies are a perfectly logical test of the hypothesis that fraud was responsible for both variance in swing and variance in redshift.

However, it rests on certain assumptions, which Kathy correctly challenges, and which I have also challenged.

If fraud was targetted in particular precincts, it might affect the Bush's vote share but not show up as a correlation. However, this would also limit its explanatory power regarding red-shift. I'm trying to figure out how much.

[sean in iowa]

57. umm, page 3 of your link contains that awful "misstatement"

In fact, it is the same page I kinked to earlier, which you acknowledged to contain an erroneous statement. Please cite a quote on that page which correctly states the hypothesis!

[Febble]

58. Sorry, we are at cross purposes here

it's difficult responding on different threads. I agree, that the hypothesis as stated by ESI is not correct.

However, it is not the hypothesis they tested. By that I mean that the analysis they did was not a test of that hypothesis. It was an test of they hypothesis I stated.

This happens more often than you'd think in statistics. People do a statistical test, but they don't correctly state the hypothesis that is tested. In this instance it's a fairly subtle error, and I certainly didn't notice it, but then I simply knew what they had done.

The null hypothesis (which is what a statistical test usually tests is)

That PLD is not greater where Bush's vote-share, relative to his vote share in 2000 is greater (or less where it is less) - which is what you pretty well stated.

Now I do not have the statistical results of the ESI finding as they have not yet published their actual report (I gather it is due out shortly). But I do have the statistical results of the Mitofsky finding (and it is deducable from the information given) which is that "the null is retained". In other words we cannot conclude that "PLD is greater where etc etc....).

The subtle difference between my statement and the somewhat careless ESI statement was that it is di.....hang on.....!!!

The ESI statement is OK - it's Kathy's bracketed interpolation that causes the problem:

Finally, ESI examined whether the proportions of the vote that Bush received in each precinct, in 2000 and 2004, were related to the difference between the reported vote and exit poll results for those same precincts. If sytematic fraud or error in vote counting occurred in 2004 but not in 2000, 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 the exit poll in those precincts.

Kathy's interpolation in bold.

The point is that the hypothesis is fine as long as it is not proposing that fraud would have resulted in Bush doing better than in 2000. Just doing "better" is fine.

It's a straight correlational analysis. Not complicated at all. Does Bush do better in precincts where PLD is more redshifted? No, it appears not. Null is retained.

[sean in iowa]

66. "Does Bush do better in precincts where PLD is

more redshifted? No, it appears not. Null retained."

Thank you. Blessed clarity. I will focus on that argument, setting aside for now the question of what ESI was or was not up to.

By the way, I know you are responding to different people here, and I can appreciate the difficulty.

So grant the results reported in your concise statement: Bush does not do better (in 2004, vs. other precincts in 2004) where precinct-level discrepancy is higher. Grant also that this is true of all 1250(?) precincts in the exit poll nationwide.

I cannot understand why this rules out fraud as an explanation for the overall exit poll discrepancy. The plausible confounding variables are powerful. Two of them:

1) As you observed elsewhere in this thread, the fraud being densest in large, high-turnout precincts where Bush was not doing well at all, but where fraud gave him a lead in the official results, or simply where fraud made him do a lot better than he actually did. It's the vote totals that matter, statewide for the Electoral College and nationally for the crucial legitimacy of the popular vote.

2) The fraud simply being done unevenly, but done wherever the perpetrators had the ability to alter the results. If you throw in pie-altering operations (shaving/spiking turnout separate from alteration of the candidates' percentages), along with percentage alterations, and use the two methods to varying degrees across jurisdictions (here we'll give Bush a tiny bump in his vote share, but give him a nice turnout bump, there we'll give him a big gain vs. his actual vote share, but increase the turnout a little, and over there we might do both to moderate degrees, and over there...), and if it was done with a watchful eye on the results of 2000 and the patterns of Democratic and Republican strength, deviating to Bush's advantage but not grossly out of sync with the expectations of Bush's and Kerry's performances in a given area, then what would the correlation of redshift to PLD look like? I'm no mathematician, but I doubt this would give you the correlation you are looking for in the exit results. I think there would be too much noise. What you would, assuming a good exit poll, would be an exit poll which showed one candidate winning, overall, and an official result showing the other candidate winning.

And as for the likelihood of the latter scenario, I'm no criminal mastermind, but if I could influence the tabulation all over the nation (and we agree our hypothetical thieves could) and wanted to steal an election, that's how I'd do it.

[Febble]

67. Well I'm working on this

As a straightforward conclusion, the finding does appear to rule out fraud as an explanation for the discrepancy.

IMO the most likely kinds of fraud, given that finding, is going to be the kinds of fraud that does not affect the exit polls (and there are several).

Vote-switching fraud would affect the exit polls, and I am currently modelling various ways of it being targetted that might fail to produce a positive correlation between PLD and swing, but they do tend to leave some other noticeable fingerprint on the data (weird distributions, mainly - and also still requires that a least some of the exit poll discrepancy is due to polling bias). So although it is strictly true to say that it doesn't "rule out fraud as an explanation for the discrepancy" it certainly places pretty strong constraints on it. But I should probably retract that blanket statement.

Although I do so rather in the spirit of one who says "no, I can't rule out alien abduction as a reason for my missing car keys, but the more likely explanation is that I yet again left them in the garage door".

I exaggerate slightly. And truly, I am working on this algorithm. But either I'd make a poor vote-switch hacker or it's damn difficult.


Pies, on the other hand, I can deal with. Pie-shrinking is certainly compatible with the plots, and a far more elegant hypothesis given the constraints. Pie shrinking could mean the election was stolen, and that the early exits were right be sheer fluke. Which might sound unelegant, but actually is more parsimonious than most hypotheses I've tried.

Of course the other hypothesis was simply that fraud was targetted in Ohio and other swing states (NM) in various ways, all below statistical radar.

I'm probably not doing your post justice - I'll come back in the morning!

[sean in iowa]

68. Two more thoughts:

1) Your statement that vote-switching fraud would affect the exits is absolutely correct. But you are looking for consistency in how they afftect the exit poll, and if vote-switching was done to varying degrees as a partner to varying degrees of pie alteration, you would not see a precinct-to-precinct pattern to redshift!

2) I propose a third confounding variable: Bush does best in 2004 in areas where he did very well in 2000. Assume that there is fraud in the areas where he is doing best. If these areas are areas where he had kicked ass in 2000, and if the exits are good, then some vote-switching fruad will not produce greater redshift because the exit is already going to show Bush doing well.

We know the above is true at the county level in Ohio: the only counties in which Bush gained 4 percentage points or more on 2000 are counties in which he won 60% or more in 2000: Auglaize, Belmont, Clinton, Darke, Lawrence, Mercer, Miami, Paulding, Shelby, and Van Wert. Dave Leip's election atlas now requires membership for county-level data for 2000. 2000 and 2004 official results here. I will e-mail you later this week the 2000-2004 table I made from Leip's data.

We need to know if Bush is doing best at the precinct level where he had already done very well in 2000, and what the exit numbers were.

[Febble]

69. Hang on

vote switching fraud has to produce redshift relative to whatever error there was in the poll anyway. So how well Bush is doing won't make any difference, unless that in itself determined whether fraud occurred.

And "how well Bush is doing" isn't measured (in these analyses) by his vote share in 2000 - but in how far he increased his vote share.

More later.....

[sean in iowa]

80. And "how well Bush is doing" isn't measured (in these analyses)

by his vote share in 2000- but in how far he increased his vote share."


Indeed, that was my understanding of the ESI hypothesis. Then you made these statements at the close of an earlier post:

The point is that the hypothesis is fine as long as it is not proposing that fraud would have resulted in Bush doing better than in 2000. Just doing "better" is fine.

It's a straight correlational analysis. Not complicated at all. Does Bush do better in precincts where PLD is more redshifted? No, it appears not. Null is retained.

Which I took to mean that Mitofsky, as opposed to ESI, was looking at 2004 results without respect to gains on vote share- in other words, is Bush's share of the vote higher in areas of higher redshift? This is a straightforward reading of your language in the above boxed quote.

Nowhere did I use absolute, as opposed to relative, 2000 results, to describe "how well Bush is doing."

So the whole kaboodle, both ESI and Mitofsky's analyses, is about expecting greater redshift in areas of greater swing.

How is this hypothesis any less vulnerable to the three confounding variables I have proposed. The first you have acknowledged as a possibility. The second (a varying use use of pie-alteration with vote-switching) you vaguely responded to by saying it would produce "weird distributions in the data." In what way? And the hypothesis, until you say it means something else, is that redshift would correlate with swing, not that fraud would produce undefined "weird distributions." The third (that Bush's biggest gains are in areas where in he did very well in 2000, and therefore where a good exit should may show him doing very well again, allowing fraud to produce nice gains on 2000 without increased redshift) you have not responded to yet.

And, come to think of it, your statement that I quote to begin this post, and your statements I put in the boxed section, utterly contradict other.

I am done with this for tonight. I will be online tomorrow around 8pm US CST.

[OnTheOtherHand]

84. maybe another voice...

I don't think those statements were contradictory at all, but as Febble said upthread, this correlation is hard to express in colloquial English.

If exit poll red shift is correlated with swing, that means (approximately) that

(Bush 2004 - Bush 2000)

is on average more positive (or less negative) in the red-shift precincts than in the blue-shift or no-shift precincts. (Of course the correlation could have the opposite sign....) That is, using swing as a measure, he does comparatively better in one type of precinct than in the other.

So, the vote outcome measure is swing -- which can be defined as "how far he (Bush) increased his vote share" -- but it doesn't matter whether he actually increased his vote share or not in any particular precinct, or in most precincts of some sort, or overall (as the Dopp paper oddly seems to insist). It only matters whether he does better on the measure in the red-shift precincts than in the other precincts. The swing measure is a comparison between 2004 and 2000, but the test is (loosely speaking) a comparison between kinds of precincts, to see if the ones with red shift generally have more positive (or less negative) swing than the ones with blue shift. It's time! it's type*! it's two "betters" in one!

*of shift, that is -- but one doesn't have to settle for categorizing the shift, one can actually measure it, thus rendering a nifty scatterplot. (I do think all this is easier to grasp if one looks at the pictures. Have you stared at Figure 3 in the ESI "Data" file?)

The confounding variables need some sorting out. I have to say, though, I actually thought it would be easier to overcome the ceteris paribus expectation of a correlation than it seems to be.

[sunshinekathy]

48. Another misdirection

Febble, you stated (twice, as I understood earlier) that:

"But what they actually tested, statistically, was what I stated. i.e. not that Bush would do better than he did in 2000 if there was fraud, but that he would do better than he would otherwise have done if there was fraud."

Again, such a hypothesis (above as you stated) would be IMPOSSIBLE to test because you don't know what the exact amount of fraud was and so have no data with which to test this reworded hypothesis. There were no Independent audits done to give any data for this hypothesis and you have no idea before applying your test, how he "otherwise would have done" with or without fraud.

So, your reworded ESI hypothesis is even more illogical than ESI's correct statement of its own hypothesis.

If you wanted to "correctly" reword the invalid hypothesis you've been testing, you could say something like,

"If there is vote fraud, then there must be a positive linear relationship between increase in Bush vote share and pro-Kerry exit poll discrepancy (WPD)."

But my proof clearly proves that to be incorrect. i.e. also invalid and illogical.

Your and Mitofsky's entire analyis is meaningless because it is illogical to suppose that you could learn anything about vote fraud from performing it.

[Febble]

51. Well, it's perfectly possible

to test the reworded hypothesis. It's what was tested.

It's how linear regression works. It's what the F ratio is all about. It's called Analysis of Variance (ANOVA), also called the General Linear Model. The test is whether the ratio of the explained (shared) variance to the unexplained (unshared) variance is greater than a critical value.

It's the basis of parametric statistics.

It's the way you determine what the proportion of variance in A is shared by B.

Which is what the hypothesis as I have stated it boils down to.

[Febble]

53. Second try:

Let me explain exactly how the Mitofsky version worked (it was slightly different to the ESI version, but not critically so).

Bush's vote share in 2004 was regressed on his vote share in 2000. There was a strongly linear correlation, not surprisingly. However no linear correlation in real life is ever perfect. In some precinct he did better than his "average" (represented by the Ordinary Least Squares regression line) and in some precincts he did worse.

Now, why didn't he do uniformly better? Well there would be many reasons. Maybe there was a better campaign by Bush in some precincts, maybe a better campaign by Kerry in others. GOTV efforts from both sides may have varied too. AND there may have been fraud boosting his vote-share in some precincts but not in others.

Vote switching fraud should, we agree, produce a redshift in the poll (although many other things will also affect the PLD, including, of course, sampling error). So what we want to know is: does the amount by which Bush does better or worse than his overall performance correlate with the amount of redshift PLD? If it does, we can say that some factor affected both PLD and swing - where PLD was more redshifted, swing was greater, where PLD was less redshifted, swing was less redshifted.

However, as I've said, lots of other things will affect both swing and lots of other things will affect PLD. But variance in these things will cause PLD and swing to vary independently - the variance due to these causes will be unshared.

So the statistical test done was to determine the proportion of shared variance between the residuals from the first regression and a measure of PLD. This involves a test as to whether the ratio between the shared variance (postulated to be due to a common factor, i.e. fraud) and the unshared variance (postulated to be due to stuff like different amounts of GOTV or sampling error, or other forms of polling error) was greater than a critical value (determined by the number of precincts in the analysis).

Alternatively we could test whether the best fit line through the plot (OLS) has a slope significantly different from zero. We can also compute the upper confidence limit of that slope. And it is very, very small.

Which means that the explanatory power of swing to predict PLD is very very small.

This does not mean THERE WAS NO FRAUD. It means that redshift in PLD is unlikely to be due to fraud UNLESS the assumptions behind the hypothesis are violated - that there was some complex non-orthogonal relationship between fraud and swing. Like fraud being targetted in precincts where Bush was doing badly. Or an algorithm that cut fraud in if Bush's vote share dropped below a certain value.

I'm still trying to figure out the implications of this, and I'm not sure I've got it all right.

But that's what I believe we should be working on together. In what ways are various forms of fraud compatible with what the ESI analysis and Mitofsky's analysis show? And is it doable? Plausible? Can we find direct evidence of it?

So I would strongly argue that we do not dismiss these analyses as ILLOGICAL - but consider exactly what they are telling us (as you do in the latter part of you paper) and what implications it has for the fraud hypothesis.

[Febble]

59. Correction to my own post

ESI's hypothesis is fine. It is your interpolated brackets that falsify it.

As long as they are not saying that fraud will result in Bush doing beter than in 2000 their statement is true, and the equivalent of mine.

Apologies to ESI, and to you, for confusing matters.

[sunshinekathy]

52. Updated - easier to understand, version of math logic proof

I just added examples composed by Jonathan Simon in the "Over-view of Math Logic Proof of the Invalidity of ESI's Hypothesis" section at the beginning of the paper.

This should make it easier for folks to follow the math logic proof.

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

[Febble]

54. Well, if you are updating

you might correct the misquote in footnote 4.

Source here:

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

[autorank]

56. Bless your heart Kathy! This is great. ESI is exposed for what it is...

...an apologist front for the Cheney-Bush cabal, one of those pseudo scientific organizations which are used by those who steal power to make things look nice and tidy (when what we really need is "Tidy Bowl").

Here's my formula: ESI = AEI (writ small)

You have sacrificed much but the products are awesome.

Thank you!!!! (btw The paper is very readable and beautifully edited.)


The walls are crumbling, the great lies are coming undone, the TRUTH is waiting, ready to expose the ultimate illegitimacy...TWO STOLEN ELECTIONS.

[sunshinekathy]

60. What is AEI?

Thanks autorank. I've seen your mathematical treatsies, although I haven't had time to study them sufficiently.

[autorank]

64. AEI=American Enterprise Institute (Neocon central)

The folks that brought us the the Iraq War, the vile Bell Curve research riddled with racism, and much much more. It's a haven for hacks who can't make it in real academic settings and who will say what the Neocons want under the aegis of an "think" (sic) tank.

I don't have any mathematical treatises, but I've explained a few. This is what you have yet to see (it's worth the read).

The Unanswered Question: Who Really Won in 2004?"

also found here, with many additional mathematical delights (you'll love this one!

When things get really tough, just remember Patrick Fitzgerald was a math major!

[Febble]

65. If that's what AEI is

then your equation of it with ESI is, I would have thought, simply libellous.

[Land Shark]

72. The laws of libel are 10 times stronger in Great Britain than in the USA!

[autorank]

76. Is the truth a problem around here now; or is it free speech;) AEI is:

From People for the American Way's Web site
http://www.pfaw.org/pfaw/general/default.aspx?oid=4456


Right Wing Organizations

American Enterprise Institute
1150 Seventeenth Street, NW
Washington, DC 20036
www.aei.org and www.taemag.com (American Enterprise Magazine)

Established: 1943
President/Executive Director:Christopher DeMuth
Finances: $24.4 million (2000)
Employees: 50 resident scholars and fellows
Publications: Monthly newsletter, dozens of books and hundreds of articles and reports each year, and a glossy policy magazine, The American Enterprise.
American Enterprise Institute’s Principal Issues:

# American Enterprise Institute (AEI) is a think tank for conservatives, neoconservatives, and conservative libertarians.
# Areas of interest include: America’s “culture war,” domestic policy and federal spending, education reform, neoconservatism, affirmative action, welfare reform.
# President George W. Bush has appointed over a dozen people from AEI to senior positions in his administration. AEI claims that this is more than any other research institution.

American Enterprise Activities:

# AEI sponsors and participates in debates and lectures on many issues.
# AEI scholars have testified before Congress on a variety of issues.
# Several AEI scholars have written articles in favor of government censorship of the arts.
# Scholar Michael Novak has argued that prayer belongs in public schools and that it doesn’t violate the establishment clause.
# AEI scholars have advocated federally-funded school voucher programs.

AEI's Background and History:

# Most of AEI’s Board of Directors are CEOs of major companies, including ExxonMobil, Motorola, American Express, State Farm Insurance, and Dow Chemicals.
# Big donors include the top conservative foundations, including Smith-Richardson Foundation, the Olin Foundation, the Scaife Foundation, Lynde and Harry Bradley Foundation.
# Corporate supporters have included: General Electric Foundation, Amoco, Kraft Foundation, Ford Motor Company Fund, General Motors Foundation, Eastman Kodak Foundation, Metropolitan Life Foundation, Proctor & Gamble Fund, Shell Companies Foundation, Chrysler Corporation, Charles Stewart Mott Foundation, General Mills Foundation, Pillsbury Company Foundation, Prudential Foundation, American Express Foundation, AT&T Foundation, Corning Glass Works Foundation, Morgan Guarantee Trust, Smith-Richardson Foundation, Alcoa Foundation, and PPG Industries.
# Kenneth Lay, CEO of Enron, was until recently on the board of trustees of American Enterprise Institute. Other famous former trustees include Vice President Dick Cheney.

AEI Fellows and Scholars List :

# Lynne Cheney, wife of Vice President Dick Cheney
# Newt Gingrich, former Speaker of the House
# Robert Bork, failed Supreme Court nominee
# David Frum, a presidential speechwriter for President Bush, contributing editor to the right-wing magazine Weekly Standard
# Christina Hoff Sommers, anti-feminist crusader, author of “Who Stole Feminism? How Women Betrayed Women”
# Charles Murray, author of The Bell Curve, a book that asserted inherent intelligence differences between the races
# Ben J. Wattenberg, host of PBS weekly show “Think Tank”

[Land Shark]

78. Truth is a defense, in both countries as far as I know.

The beautiful and elegant thing about a defamation case in the U.S.A. at least is that, by placing reputation in issue, all kinds of testimony concerning reputation that would in other types of cases be deemed irrelevant, comes into the record instead.

But I think Febble's point assumes AEI is bad and questions its application to ESI.

[autorank]

79. Well, my analogy was highly provocative I guess. AEI bears a good look.

[Bill Bored]

83. How about ESI = PNAC?

OK just kidding. At least contribution to ESI are tax deductible!

(And I don't think ESI ever outed a covert CIA agent either.)

[autorank]

85. How about Bill Board = Grand Master, Priory of Sion

:evilgrin:

[Bill Bored]

95. Nah, I'm more into Freedom Fries, Auto (sans the Fries). nt

[autorank]

103. You always seemed like an Order of the Golden Dawn guy to me.

Allister Crowley

:evilgrin:

[Land Shark]

87. 'tis true, ESI is keeping *their* CIA agents under wraps! : )

But disclosing that 10% of their funding is from vendors.

[autorank]

90. Those agents, part of the "remote viewing" program.

That CIA's smart. They close it down but really just move it "off campus." "I see fake ballots..."

[Bill Bored]

94. NO! Have they? For REAL?

If so, now they've gone too bloody far!
Do you have a link for this LS?

I know some folks who'd be very interested to know about that (but they're so hip they probably know about it already). But, just in case....better to be safe than sorry!

(Unless of course you're talking about soft drink vendors or something.)

[Melissa G]

96. Yeah Land shark, are you pulling our collective leg or do you have

a linkety link for us on this 10% vendor scoop?

[Bill Bored]

97. I wonder if it's the same vendor(s) with whom they've signed NONDISCLOSURE

AGREEMENT(S)?

[Melissa G]

98. that would be Diebold, correct?

[Bill Bored]

99. Well, it ain't Coke and Pepsi!

But actually I think it's ES&S.

HOWEVER, we should NOT let this influence our judgment regarding their opinion of what caused the Exit Poll Discrepancy! Their work should stand on its own (whether they can explain it or not)!

[autorank]

105. Spoken like a true empiricist.

I do apologize for outing you in the post above (Golden Dawn) but I think you should embrace your heritage.

[Land Shark]

111. I haven't looked into it, therefore I have "no evidence" of CIA agents

The part about vendor funding of ESI is true according to John Gideon at Votersunite, however.

I think when people make claims about "no evidence" they should identify what search they've done on the subject and what evidence has been directly reviewed:

Answer: re specifically the existence of covert agents at Voting Vendors:
Search: None.
DOcuments reviewed: None.
Strength of "no evidence" conclusion: Therefore, none.
Motivation to find or expose: Reasonably good (but irrelevant in light of negative answers above)

Most of the people claiming "no evidence" for this or that fail the "motivation to find or expose" test, if not all of the above tests.

Now that being said, I'm told that there are certain (not only software) but hardware backdoors built into computer chips for "national security" purposes that our spooks have access to. So it's a reasonable inference that there are methods available to a person or persons with enough knowledge or access. Certainly fly by night's info on a recent thread is that vendors ADMIT that they have two way wireless communication capabilities not sought or needed.

[FogerRox]

86. Our dear friend Mr. Bork--- dint he start the Federalist Soc. after he got

turned back from the SC?

Melissa-- dont forget the cult of personality--you go girl

[autorank]

88. Cult of personality disorders? Multiple cults! Mini cults?

Damn, I'm headed over to read some Justin Ramone or whatever his name is to get the full story!

[Febble]

107. See below

I approve of American laws regarding free speech.

It was a figure of speech. I forgot Americans took litigation so seriously. We may have stronger libel laws, but we don't sue so much. And actually in many ways, our libel laws make for a good independent press - even the gutter press tends to check its facts - truth, in the UK as in the US, being a defence against the charge of libel.

"Hyperbolic" was the term I should have used.

ESI bears no resemblance to the organisation described in this post.

[Land Shark]

110. Situation is critical now in US regarding free speech, because Dan Rather

case proved that even if you have an expert witness and other backup, if it turns out that DOCUMENTS you have have backed up by an EXPERT and witnesses just MIGHT WELL be fakes, you may lose your career even if you are the top anchor at a major network.

I think that sends a message in the strongest possible way to self-censor unless one reaches near absolute certainty, which is rarely if ever possible.

[Febble]

104. Good for the USA

although we manage nonetheless to have a more independent media.

But it was a figure-of-speech for an equation so absurd there should be a law against it.

I forgot Americans were so literal about litigation.

(We may have stronger libel laws, but far fewer people ever sue anyone).

[sunshinekathy]

93. New Version - Jonathan Simon & David Griscom names added to Proof Paper

An updated version (slightly).

David Griscom, physicist, election researcher in Arizona, and speaker at some election conferences has read and signed off on the paper.

Jonathan Simon, the person responsible for downloading and saving all the screen shots of the pre-adjusted actual exit polls on the night of November 2nd, who gave them to Steve Freeman to analyze, contributed a few sentences to the paper and some math logic that should make it more readable.

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

I want to be clear, this is a proof that the analysis by Mitofsky, ESI, and Liddle that they've been claiming rules out or weakens the vote fraud theory, is BUNK, meaningless, and shows nothing, one way or the other, about vote fraud.

i.e. any relationship between Bush vote share increase from 2000 to 2004 and exit poll discrepancies can co-exist with vote fraud.

Take our paper to any mathematician who knows logic and ask them to verify its correctness.

Valid analysis of vote fraud is done by other methods, but not via the methods that Mitofsky, ESI and Liddle have been performing since June.

NEDA will be coming out with a valid analysis of Ohio precinct level exit poll data on November 2nd.

Best,

Kathy

[foo_bar]

100. it seems more like a fundraising letter

If steps to implement it are begun today, it could be operable by November 2006.

Gee, that's always the conclusion!

The proof-in-name-only at center is a fallacy known as setting up a straw man:

The ESI hypothesis, proven herein to be logically incorrect, is stated in an ESI paper as follows.
"If systematic fraud or error in vote counting occurred in 2004 but not in 2000, Bush would have done significantly better in those precincts in 2004 2000], and we would see larger differences between the reported vote and exit poll in those precincts ."

So the ESI hypothesis is one sentence, a substantial portion of which is in brackets. Yet "an ESI paper" isn't even named much less footnoted, so the burden's on the reader to discover the source of the quote, the source of the context you added, and the source of the context you eliminated (no bibliography either; the only footnote on J'accuse page 1 refers to another (questionable) USCV report). Same goes for the apocryphal Mitofsky quotations.

Needless to say, that doesn't pass for science. The paper seems to have academic pretensions (judging from the layout, which is quite nice btw), but it doesn't come close to meeting the standards which govern conduct off the internet:
http://www.ucl.ac.uk/Library/CitationPlagiarism.doc

But you already know something's wrong when the coauthors are the peer reviewers.

I would hesitate to thank those who are throwing out hastily done analysis without checking it first, and therefore misleading the American public into thinking that there is no evidence of vote fraud in the 2004 presidential election results.

Ironic, no?

[sunshinekathy]

101. Donate & Take it to a math professor - if you can't understand it

You obviously do not follow math logic (and not everyone does), so please take it to a math professor whom you respect and ask him/her to explain it to you and they will verify for you that it is a valid correct proof.

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

You proclaim that the ESI paper is not cited in my math logic proof when it very clearly is cited and named as are all its authors and its URL is given, and you can always use google to find it too, if you can't read well.

Do you want to ensure that the correctly elected candidates are sworn into office following elections or not?

If you do want democratic elections restored, please donate:

http://electionarchive.org/fairelection/donate.html

There are many links from the donation page above that explain why donations are needed.

Building a national election data archive is a huge project that cannot be done without many staff and funding, and it is the only tool that would ensure that correctly elected candidates would be sworn into office following elections by Nov 2006 IF it were funded.

Do you think that democracy is worth restoring - even if projects need to be funded to restore it?

The cause of democratic elections is worth funding -- every other issue depends on having the right to have our votes counted accurately.

Perhaps for folks like you, we should have listed the academic credentials of the author and contributors and reviewers, but I am just one human being trying, not very successfully, to work full-time without pay, over 100 hours per week to try to do the work that five or six people should be doing, to implement the only project that is practical, if people would only fund it, for ensuring that correctly elected candidates are sworn into office following the 2006 election.

Read more about it here: http://electionarchive.org

To tell you the truth, I've been shocked and dismayed that folks have not donated already to get this project off the ground. I've been dismayed and frustrated that more folks don't see the urgency of restoring democratic elections in America, and you would think that folks who have not given up their full time jobs like I have, to work full time on this issue could at least donate $10/month to save democracy. If only 250 people did that, we could at least hire one full-time programmer.

I personally believe that lives all over the planet, and the future of civilization itself depends on us restoring democracy, and so you and everyone else too, should donate to the cause of restoring it.

http://electionarchive.org/fairelection/donate.html

[Febble]

102. I've got a better suggestion

Take it to a statistician. A real one.

Not a physicist, or an attorney.

Someone who can explain the nature of a regression equation, the nature of a residual, and the nature of variance.

And "check" the "logic" of the paper before you "rush" it to the "American Public".

I also suggest you correct the citation errors (can it be that hard to copy and paste?); address the hypotheses that have actually been made, not the ones you have inferred (hint: it is not difficult to figure out the hypothesis tested by a statistical test - try a text book on regression) and provide links to and/or sources for all alleged quotations. If you can't find links or sources, I suggest you contact the people concerned for on-the-record confirmation of the hypothesis they claim to have tested, and the inferences they claim to have made.

In short, check that the people you are allegedly "refuting" actually hold the positions you "refute".

To do otherwise, is, IMO, to produce "shoddy" and "irresponsible" work that is likely to "mislead" the "American public".

I expect no less "due diligence" from one who demands it so vociferously from others.

And yes, I do seem to have stopped being "polite". I'd hate anyone to think I was a "con artist".

[sunshinekathy]

123. Wrong - statisticians do not always understand math logic

Many statisticians, such as Liddle, and many others do not have mathematics degrees. They only have degrees in some other area, and some training in the technical aspects of statistics.

You need to take this math logic proof to a real mathematician, who has had training in ALL areas of math, not just the techniques of statistics.

It is very easy for some people to obtain training in statistics without obtaining the logic skills that they would need to make sure that their own work is always logically correctly applied.

You need to take this proof to a person with mathematical training in math logic, who is sure to understand it.

Many people learn statistics without learning all the other mathematical background that makes them a little better at making sure that they always apply their own mathematical techniques only when appropriate and logical to do so.

[foo_bar]

106. does high school logic require a math professor to explain?

Therefore A->(B^C) is not a logically correct inference and one cannot reach ESI's final conclusion ¬A, that vote fraud is ruled out.

Except ESI's actual conclusion is:

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 results

So your Not A is hyperbole, to put it charitably.

Here's the ESI paper you didn't link to (unlike the copious USCV footnotes):
http://www.votewatch.us/Members/stevenhertzberg/report.2005-07-19.7420722886/report_contents_file/

[Land Shark]

109. I think the USCV is fair about how the ESI report is ACTUALLY used

in the press and by many of its supporters. And that outcome is, to cast cold water without particularly moving anything forward. Put another way, the outcome is "no story here on fraud".

Foo-bar, the ESI report definitely is linked in the report.

Febble if you are referring to Moi when referring to giving the paper to an "attorney" or "physicist" as being less adequate than a statistician, that would be inaccurate since I've not been given the paper any more than anyone else on this thread.

[Febble]

113. Not you!

But upthread Kathy says that Jonathan Simon, who I believe is an attorney, is now a contributor.

Yes, ESI is linked. However, the statement of their hypothesis as quoted by Kathy contains interpolations, presumably by Kathy. It is the interpolations that make the hypothesis patently false.

As it is written their report it is a perfectly valid hypothesis.

Kathy, elsewhere on this thread, has commented that the hypothesis as it actually stands is not testable.

This would appear to reveal a fundamental ignorance of inferential statistics. I suggest she runs it by a few statisticians.

[Land Shark]

114. Oh, I forgot Simon is an attorney, but I've no problem with his skills

Many of the best attorneys come from at least backgrounds in math or science, since the training in analytical science is helpful in law. Limited disclosure: LS has a bachelor's degree in biology.

[Febble]

116. Well I simply have a problem

with the math skills of anyone associated with that paper. The more I think about it, the crosser I get.

I'm so used to the flying brickbats aimed at me that I scarcely notice them. But abuse of logic is something else.

Kathy "refutes" the nonsensical proposition:

If there was fraud, then Bush will do better than he did in 2000.

This is clearly hooey. Fraud might be the only thing scraping him off the ground.

However, the proposition:

If there was fraud, then Bush would do better than he would have done without fraud.

Is self evidently true. Unless of course the fraud was designed to benefit someone else, in which case they would do better. But let's take it that fraud is pro-Bush fraud.

So if we now propose that:

If there was fraud, and if the fraud was responsible for redshift in the exit polls, Bush should do better than he would otherwise have done in precincts with higher redshift than precincts with lower redshift, or blue shift.

Kathy maintains up-thread that this isn't testable. It's perfectly testable. However, lots of things will affect how well he does. One big determinant will be how well he did in 2000. In a precinct full of his 2000 supporters he will do better than in a precinct full of Gore supporters. So we can start by removing that determinant from the equation. We are then left with the proposition:

If there was fraud, and if the fraud was responsible for redshift in the exit polls, Bush should do better than he would otherwise have done, given his performance in 2000, in precincts with higher redshift than precincts with lower redshift, or blue shift.

Again a number of things will determine this, such as the success of local GOTV efforts (which will affect the performance of the candidtates). There will also be variance in polling accuracy due to reasons that have nothing to do with fraud (sampling error, for one).

Now, the way the relevant statistical test works is that we look at the ratio between shared variance (variance shared between Bush's performance in 2004, given his performance in 20004, and PLD) and unshared variance. Shared variance will, according to our hypothesis, indicate fraud. Unshared variance will indicate variance in both variables due to independent factors (eg. GOTV, sampling bias).

So the hypothesis NOT refuted by Kathy but actually TESTED by ESI and Mitofsky is perfectly testable - and the answer is that the unshared variance completely swamps the shared variance (F ratio laughably small in the case of the Mitofsky analysis).

Now we can discuss, as we will, and have been, the implications of this for various forms of fraud. And it may even turn out that there is a mechanism whereby fraud is not only compatible with the analysis (clearly it is - there are many forms of fraud that won't affect exit polls anyway) but which might allow fraud to produce redshift. But I have to say, despite much effort, I'm finding that latter thing elusive.

[Land Shark]

118. Here's how the hypothesis apparently SHOULD read:

Febble said the correct statement would be:
"If there was fraud, then Bush would do better than he would have done without fraud."

This is certainly not even close to true unless we add:
"If there was fraud **by person(s) favoring Bush**, then Bush would do better than he would have done without fraud."

But even that's not true unless they are competent, so:
"If there was fraud **by person(s) favoring Bush**, (**who were competent enough to achieve the objective desired**), then Bush would do better than he would have done without fraud."

But now, clearly stated, we have nothing but a meaningless tautology.

The only possible way out of this box (it seems to me right now) is if you show precisely why what Bush would have done without fraud is a meaningful statement. That seems unknowable without at least first provisionally accepting that fraud occurred and its amount. You attempted some complicated variance explanations that claim they get around this difficulty but they make no sense to me.

[OnTheOtherHand]

115. good point on the link (although there is a problem even there)

foo probably missed it because the ESI report is quoted on page 1, alluded to on page 2, and finally referenced on page 4. That's not quite the way citations are supposed to operate. But that is merely sloppy.

I don't think the USCV paper can be defended as "fair" on the grounds you suggest. The USCV paper, if we are to take it seriously, is not at all about how the ESI report is "ACTUALLY used," but rather about what the ESI report purportedly says that is allegedly logically fallacious. The distinction between intellectual and political criticism is one that empiricists generally work very hard to maintain. If someone prefers the USCV report to the ESI report not because the arguments are better, but because the politics are better, so be it. But the USCV report does not purport to be a political critique. If it is to be defended as fair, it should be defended on its own terms.

[sunshinekathy]

124. That is NOT ESI's conclusion - read their paper

And math logic DOES require college level training.

If you learned math logic skills in high school great. But I don't know anyone else who did because math logic courses are not usually offered in high school as they are in college.

You keep misstating the facts.

Why? What is your motivation?

[OnTheOtherHand]

108. remarkably undisciplined

I want to be clear, this is a proof that the analysis by Mitofsky, ESI, and Liddle that they've been claiming rules out or weakens the vote fraud theory, is BUNK, meaningless, and shows nothing, one way or the other, about vote fraud.

i.e. any relationship between Bush vote share increase from 2000 to 2004 and exit poll discrepancies can co-exist with vote fraud.

There is a huge difference between saying that any relationship between swing and exit poll red shift "can" co-exist with "vote fraud," on the one hand, and saying that "the analysis by Mitofsky, ESI, and Liddle... is BUNK, meaningless, and shows nothing." Do you really not understand this?

The idea of a logical refutation of a correlational analysis is deeply weird. As I wrote on the 22nd, "Just as 'correlation doesn't prove causation,' lack of correlation doesn't prove non-causation" -- hardly an original insight, but rather Data Analysis 101. Your paper seems to add very little to that point beyond some very misleading representations of other people's arguments. It is a wasted opportunity.

[Land Shark]

112. OTOH "Deeply weird" takes the unprofessional cake, but

you could be free to issue your own brief report stating that while you do not concur that the entire analysis has been proven logically flawed, your analysis indicates that fraud is entirely possible in the 2004 election and not ruled out by the ESI analysis.

I trust you would put the strongest justifiable terms in the place of "entirely possible".

This would seem to advance the cause of science/stats as you see it, and be accurate, and keep the press from misunderstanding the import of ESI.

[OnTheOtherHand]

117. no, that would be special pleading, but...

First of all, while I wouldn't write "deeply weird" in a journal article, I think the language is appropriate. To use symbolic logic to declare a correlational argument "BUNK" is, well, deeply weird. That sort of thing wouldn't make it into a professional journal, so it is hard to know what else one is supposed to say about it.

Land Shark, under the circumstances, what you propose would actually be highly unethical in my opinion. The USCV work does not meet basic professional standards, and should not be indulged. The depth and breadth of its flaws impose limits on how nuanced one can be in response. This is a source of great frustration and pain for me.

[Land Shark]

119. The central apparent problem we have here, I'm not sure OTOH recognizes

which is (on the assumption that OTOH is largely correct) we have a huge "unknowability" problem given limited data. Yet, at the same time, democracy must be defended (whether under attack, or not), so there has to be analysis given what's there. Thus, those who rely on or point out unknowability simply support the status quo unless these claims are joined at the hip with disclosure demands at all reasonable times.

Another debate: I think one simply MUST, in the "ethical" world, be aware of how one is or isn't being "used" in the political sense. Otherwise, one will enter into all kinds of unfortunate relationships, even if on some technical level, they seem ok at the time.

[OnTheOtherHand]

120. I don't disagree

Indeed, what I perceive as the screeching political irresponsibility of certain arguments troubles me at least as much as their IMHO intellectual irresponsibility. Anyone who loudly makes a poor argument for fraud in 2004 makes it that much harder for others.

Now, in other circumstances we have hashed out what is wrong with the ESI brief -- opinions varied somewhat, but no one (including me) liked it very much.

ESI did couple its report of a pertinent empirical finding with the observation, "The fact that there is debate over this at all shows that we need elections to be more transparent, more accountable, and more auditable. The public has a right to know exactly how elections work, and to verify for themselves that the voting and the counting is done right." (Page 4 of the brief.) I do not believe it could fairly be said that ESI is, at least in this context, among those who "simply support the status quo." In Philadelphia, as far as I can tell, the strongest call for change came from Mary Batcher, not from Steve Freeman.

(And I note again that what ESI presented was an important empirical finding; it wasn't mere handwringing about "unknowability.")

[Land Shark]

121. It's got to always be joined at the hip with the calls for disclosure

or else it is at high risk of being used/abused.

[sunshinekathy]

125. Any analysis based on an invalid hypothesis IS Bunk

Obviously many people (yourself included) do not understand that ANY analysis which is based entirely on a faulty invalid illogical inference and hypothesis like Liddle, Mitofsky and ESI did is meaningless bunk, so that is why I spelled it out for you explicitly.

This is just absolute provable fact. There is a field of science called mathematical logic that you, and many others, have obviously not studied because it is a college-level optional field of study.

Get over it and stop trying to shoot the messenger.

[davidinalameda]

126. Locking

This thread has turned into flame bait

Please try and keep discussions on an adult level without resorting to name calling and little snarky comments directed against others.