Kathy Dopp's Most Recent Post Concerning Lizzie Liddel's Criticism of USCV

[davidgmills]

Because it Deserves it's own Post:



sunshinekathy (12 posts) Sun May-01-05 12:47 PM
Response to Original message

24. US Count Votes' Rebuttal & Plea for Support of Election Data Archive


Bruce O'Dell, VP of US Count Votes, has created a simulation model of Lizzie's algorithm. It indicates that her model requires a "participation by partisanship" profile that is -totally- inconsistent with E/M's empirical data.

Lizzie's model requires -greater- exit poll participation in High-Kerry precincts than in High-Bush precincts. E/M's data shows quite the contrary.

If you reproduce Lizzie's results, and delve down into the data, you face a fundamental contradiction with E/M's "participation rate by precinct" empirical data.

Lizzie (and E/M's) "uniform response bias" hypothesis simply does not fit the facts. Period. There is simply no way to reconcile a "uniform response bias" hypothesis with E/M's "participation rates by partisanship" data.

US Count Votes will eventually publish the raw data and my simulator source code, to allow anyone to confirm my assertions.

For Bruce O'Dell's professional qualifications to simulate mathematical models, please refer to www.digitalagility.com/Odell_home_page.htm
35 years after writing his first Basic program, Bruce has reached the peak of his profession.

He is currently engaged as the lead architect at an enterprise security project at one of the Fortune 20 ("20" is not a typo). He is in charge of the technical aspects of security at one of the twenty top public corporations in America, and for the last 15 years, I have focused a major portion of my consulting practice on formal modeling of complex systems.

US Count Votes needs funding NOW to hire full-time programmers, system administrators, liason with the 33,000 election offices in America, statiticians, etc. if it is to complete its project to audit elections to have a system in place by Nov 06 to ensure that the right people are sworn into office in Jan 07. This cannot wait. The technical systems US Count Votes is building for its National Election Data Archive project are huge and cannot be built in six months prior to the Nov 06 election.

Every other issue that the Democrats care about will be resolved as soon as we audit elections to ensure that the right candidates are sworn into office in Jan 07. US Count Votes' needs 1/2 million dollars minimum for the staff to accomplish its database systems (which require about 1000 database tables - 15 per state and some shared tables). Please help out by donating or helping us to raise funds:

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

This project is the Democrats' best hope for restoring Democratic control of Congress in Jan. 07.

Best,

Kathy Dopp
http://electionarchive.org

[ibid]

1. Will the media turn interested and honest if we show more proof of a

stolen election.

I agree this is importany work that deserves support.

But without a change in the right wing bias of the national media so that headlines, spin, and indeed story selection becomes a little less all "GOP is great/Dems bad" all the time - with GOP problems buried on page A19 on Saturday - I do not see this changing the situation.

The media has no ethics these days

[Peace Patriot]

2. Wow! Well, friends, talk about a GOAL! Here it is. They need a half a

million dollars. Seems like a daunting amount, but really it's a pittance when you consider what's at stake, and the billions and billions and billions of dollars that the Bush Cartel and its friends are stealing from us and others, or that are being poured into Democratic Party and other coffers, as we speak, to pay the news monopolies for political ads, or the money we all poured into Dem coffers and the K/E campaign to keep them competitive with the Bush Cartel money machine.

George Soros? Ted Turner? Barbara Streisand? Whoopi Goldberg? Tim Robbins? Susan Sarandon? Jane Fonda? (--just read that Ted gave her a $10 million foundation before they split up). Bruce Springsteen? Eminem? Bonnie Raitt? Air America? MoveOn? Millionaires, billionaires, media savants, media stars and other powerful people with good hearts: HE-E-E-E-E-LP!!!

Okay, we need a thermometer on every election fraud/election reform web site: Freepress.org. Solarbus. Truthout. DU. Votersunite. Etc. And a common effort. Everybody's got their own fundraising problems, but this is our common need: fraud detection in '06. It is a MUST. And it just can't be done by scattered, state/local grass roots efforts, no matter how fervent. We need a collective database run by and analyzed by recognized experts. Kathy Dopp is so right about this.

How do we get this done? Ideas?

[Peace Patriot]

4. In response to ibid, above: This has nothing to do with the media...

...monopolies. This is project of THE PEOPLE--OUR government, OUR election system, OUR majority collective will, OUR demand for proper and rightful representation, OUR future. NOBODY in government OR in the news monopolies is seeing to this: the verification of our elections, and the honesty and integrity of our election system. So we MUST do it ourselves. That's what Kathy Dopp is talking about--creating a PARALLEL ELECTION MONITORING SYSTEM, independent of the one that is giving us false results, where, at the very least, we KNOW WHAT HAPPENED, where the black holes are located into which are votes are being "disappeared," and can determine courses of action based on quick and expert assessment of the facts.

It has taken six months, since 11/2/04, to put together a comprehensive picture of that election--and there are still big holes in our knowledge (for instance, a lack of precinct level audits almost everywhere). We absolutely need to quicken this process that we have all been through over the last six months. One of the things we've learned is that government AND the news monopolies are FAILING US miserably on even the simplest and most common sense election verification measures.

This is the counter-measure--and it is well thought out, potentially very well organized, and intended first and foremost as a system for informing THE PEOPLE--not the government, not the controlled news media--of what is going on, and doing so with great immediacy.

If this system had been in place in 2004, we might have had a quite different result from the efforts in Ohio and other places to verify the election and detect fraud--and we might not have George Bush as head puppet of this fascist coup today. We need the ability to instantly focus election monitoring efforts on exactly the right places, on a nationwide scale, so that we can rally local support wherever it's needed, instantly, to whatever action is needed, for election verification--counting voter sign-ins at the precinct level, securing evidence, demanding transparency, protesting, suing.

We had desperately overworked and underfunded election activists rallying on an ad hoc basis, in Ohio for instance. It was an heroic effort, in response to massive fraud that the country was not at all prepared for. Such efforts should be more routine than heroic, and should have much more pre-planning and support.

We may lose, still--as you seem to be predicting, ibid--but, by God, THIS TIME it won't be because we weren't ready for them and their damnable election theft, on OUR OWN INITIATIVE--independent of useless or criminal Secretaries of State, fraudulent attorneys general, and oblivious or corrupt Democratic Party leaders.

The "Last Stand" metaphor has perhaps been overused, but there is a lot to say for 2006 being the "Last Stand" for democracy in the United States of America. Personally, I feel that democracy is never dead if there are people who still believe in it, and are fighting for it, no matter how bad things are. Democracy is a goal, a purpose, a shining ideal. So, I would put it like this: 2006 is the "First Stand"--the opening battle, the second "shot heard round the world"--in our struggle to re-establish our democracy, and maybe even to create an improved version of it.

[Amaryllis]

28. Add Leo DiCaprio and Woody Harrelson to that list. They are both super

activits. May need some educating, but they are progressive activists.

[Febble]

3. From Elizabeth Liddle

As you've re-posted Kathy's comment, I'll re-post my response below.

But can I say first that I absolutely back Kathy's appeal for funds. The fact that we are even debating whether the exit polls or the vote count were wrong is a scandal. Elections must be both fair and seen to be fair. Also the voter suppression in Ohio must never happen again. When is Blackwell going to jail?

And Kathy is working tirelessly in Utah to prevent DREs being bought there. You must be able to audit your elections directly. Exit polls are far too crude, especially when response rates are so low.

Which takes me to: response rates:

The E-M report states that there was no significant difference in response rates between precinct partisanship categories, although they do not give details of the statistical test. However, it is true that the "differential response" hypothesis predicts that response rates should actually be lower in Bush strongholds than in Kerry strongholds - because allegedly, Bush voters don't like responding.

This is an important argument. However, statistically, a small difference is only detectable if you have enough statistical power. The predicted difference is small (in theory you should get 56% response from an all-Kerry precinct, and 50% in an all Bush district, so the maximum difference is 6 percentage points). Statistical power depends on two things - plenty of data and good signal-to-to noise ratio.

Here the numbers are small where it matters - in the extreme precincts where the differences should be maximised. There were only 90 "high Kerry" precincts and only 40 "high Bush precincts". So the data would have to be pretty clean. We know it was not, as the E-M report gives as a value called the "absolute" error, which, without going into too much detail, tells us how much noise there was in the data.

When I did my model, I compared the amount of noise in my model data with the amount of noise in the real data. I had less, and even so, statistical significance of the lower responses in my "high Bush" precincts was only borderline. I estimated that if I increased the noise so that it matched the real data, the difference would be lost in the noise. It only takes a couple of stray data points to lift the response rates at the "high Bush" end, or lower them at the "high Kerry end" to destroy the effect.

So I agree that the lack of evidence (and it is no more than that)of lower response rates in "high Bush" precincts than in "high Kerry" precincts is an argument worth considering. However, the way statistics works that what you test, statistically is a "null hypothesis". If you cannot show that an effect is unlikely to have occurred by chance you "retain the null". If you "retain the null", have not proved that the effect was not there. You simply have not proved that it was. It is the statistical equivalent of "presumption of innocence".

Unless E-M provide us with better data, I do not think we can reject the null hypothesis that over-estimates of Kerry's vote were randomly distributed amongst precincts of all degrees of partisanship.

This in itself, however, does not rule out fraud as a cause of the over-estimate. The same effect would result from under-counting of Kerry votes.

Bruce has done a great job with his simulation, and he is clearly at the top of his field. However, the point here is not the skill of the modeller, but the parameters of the model. With the greatest respect to Bruce (and we have been exchanging information very co-operatively over the modelling, for which I am extremely grateful to him), I simply disagree that his model disposes of widespread rBr, or for that matter, of widespread fraud. (Other forms of analysis may do the latter, but not this one).

This may be a matter of different philosophical approaches to data. Bruce is an engineer, my background is in the social sciences (psychology) although I also have done a bit of engineering. When data points are people, noise behaves badly. I think what we are seeing is noise, and I would argue that there is a sound statistical basis for my claim.

Lizzie

[Peace Patriot]

5. And how do you explain the virtually impossible skew to Bush in...

...the official tally vs. the exit polls in the battleground states, and other states needed to win? Is this not the more important finding, as to election fraud?

Also, what if the "Reluctant Republican Responder" was a Republican voting for Kerry, reluctant to disclose such a vote in the notoriously repressive political environment of Bush-era Republicanism? As to guesses and speculations, with little or no data to substantiate them, isn't this a much more likely Reluctant Responder than a Republican voter for Bush (and especially so in Republican strongholds)?

I think it's very telling that Edison-Mitofsky doesn't consider this possibility.

[Febble]

6. I am not sure exactly what you mean

by the "virtually impossible skew to Bush" in the battleground states, and in any case was not what I was addressing in my paper.

It is true that Ohio, Pennsylvania and Florida had larger errors than some other states, but the one of the largest errors was in New Hampshire, not a battleground state, and they also had a recount that satisfied Nader.

I would like to know whether the within-precinct error was larger in battle ground states or not. My own analyses of the WPE suggests it was not, although as I have said elsewhere, I think these analyses need to be done with a different (my) variable. Of the three big battleground states Pennsylvania was the most out of line, but not statistically significantly so. Other, less sensitive states were further from the mean.

I think you misunderstand the "reluctant Republican responder" hypothesis. The E-M analysis was based solely on votes, regardless of party affiliation. It is normally called the "reluctant Bush responder" hypothesis, as the hypothesis is that voters who voted for Bush were more likely to refuse to respond than voters who voted for Kerry. Both were reluctant - but the figures estimated by E-M are that while 46% of Kerry voters managed to avoid the pollsters (by luck or design) 50% of Bush voters did. Not a huge difference, but enough to seriously upset the predictions.

But my paper does not argue for the reluctant Bush responder hypothesis. It only argues against the "Bush strongholds have more vote-corruption" hypothesis. If you want to read my paper as evidence for much more wide-spread vote-corruption (which there has to have been if the exit polls really reflect vote-corruption), feel free. I think there is other evidence against that view, but that is not what my paper is arguing.

[TruthIsAll]

8. Your focus on "battleground states" is misplaced.

South Carolina was not a battleground state.
Neither was Georgia. Or Texas. Or Mississippi. Or Alabama. Or Louisiana.

The fraud was not just focused on winning the Electoral vote in targeted states, but also to win a "mandate". It would not look good if Bush stole OH and FL and lost the popular vote by 6 million. So even though votes were stolen in the battleground states, these could not provide the popular vote "mandate".

Bush did not win by three million. He lost by at least 5 million. Read the analysis.

[davidgmills]

9. She's entered your domain now.

The TIA twilight zone.

She may regret this.

[Febble]

16. Fine

But I was asked to account for the extra error in battleground states.

It seems to me there is some goalpost moving here.

Q: "How do you account for more error in battleground states?
A: "I don't think there was more error in battleground states."
Response: "But that means they just wanted to win the popular vote!"


Here's another one:

Q: "How come the bias was greatest in strongly Bush precincts?"
A: "It wasn't greatest in strongly Bush precincts"
Response: "That's because it was everywhere!"

I don't KNOW if there was fraud in that election. I think there probably was. It is a scandal that we don't know. It is why I have been supporting USCV's efforts to make future elections auditable.

But I think this particular paper by USCV makes unjustified inferences from the data, and my paper demonstrates why. Furthermore, I don't think Bruce's ingenious model (and he has been sharing his algorithms with me as I have been sharing mine with him) refutes my conclusions. He disagrees, as he is entitled to do. We are both numbers guys, like you, trying to figure out what went wrong.

Please read my analysis.

[davidgmills]

7. Why is an hypothesis, that is the equivalent of a presumption of innocence

be advanced as the most likely source of the discrepancy?

We have debated on this board for several days how an hypothesis, which can't be tested, but cannot apparently be ruled out according to you, be advanced as the reason for the discrepancy. Some argue that scientifically, rBr is technically not an hypothesis, because in scientific terms, an hypothesis requires an idea that can be tested.

For me, the only way to prove the rBr hypothesis is to prove the count correct thereby eliminating fraud or incorrect count as the culprit.

We know there was some fraud and we know that statistically, the count could not have been correct.

[tommcintyre]

11. David, rBr IS a hypothesis (a guess) , but can't be tested, so...

it can't be used to explain the exit poll discrepancy.

You are right, proving the count correct (or incorrect) would be very useful. However, even if we had access to ALL the exit poll data, and were able to recount ALL the votes, I still doubt we could eliminate fraud or incorrect count, since 30% of the votes were on DREs. Many have no independent record so they can only regurgitate the same results. Others have been outright erased (New Mexico). And of those that produced a "paper trail", unless the voter confirmed the output, we can't be sure the paper wasn't produced AFTER the actual vote and modified.

One other point that shows just how weak the rBr hypothesis really is: Even if it were possible "...to prove the count correct thereby eliminating fraud or incorrect count as the culprit", it still would not prove the rBr hypothesis. Other hypotheses (guesses) could also be proposed and tested. The one(s) that held up best under testing would no doubt receive the most support as the possible explanation for the exit poll discrepancies.

Here are my original posts (and the replies) to my challenge of the validity of advancing the rBr hypothesis as an explanation for the exit poll discrepancies, since it can't be tested:

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=363706#363731

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=363383#363733

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=364217#364355
<Febble has replied to this post. I am revisiting the applicable documentation, and will respond to her post by tomorrow.>

Your related follow-up thread:
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=364390

David, You say: "We have debated on this board for several days how an hypothesis, which can't be tested..." And: "Some argue that scientifically, rBr is technically not <it is> an hypothesis..."

If there are other threads/posts discussing this specific challenge to the rBr hypothesis, please let me know since I am developing a more formal argument. It would be nice to have more input on the idea. (As I asked you for two days ago (via PM).)

Your reply to my first post on the subject (first link above) was:

"You are on to something.

I don't think it can ever be tested. How are you going to use math or statistics to test what is in the mind of an unknown person?

The arguments I got to refute yours were really far fetched.

davidgmills"
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=363383#363963

It would be useful to have a better idea what your friends actually said (especially if one of them you ran this past was Rick Brady ;) ). I await your reply.


Thanks,

Tom

[davidgmills]

14. Sorry Tom

I now remember this was your line of thought and should give you credit. I couldn't remember earlier in the day who made the post and frankly was too lazy to go searching.

Also I never check my private messages. My bad as they say.

[Febble]

17. I think you guys need to read my paper

You are attacking a straw man.

I am not supporting the rBr. I am saying it is consistent with the data we have. The two are not the same.

My paper did two things: it called into question the variable that E-M used to test the rBr (which should please you guys), and it also called into question the inference made by USCV that vote corruption was greatest in Bush strongholds.

I don't even rule out vote-corruption everywhere. Please read the paper if you are going to discuss my conclusions.

And rBr is, contrary to your assertion, a testable hypothesis. I am arguing that it has not, to date, been adequately tested.

It is a more testable hypothesis than fraud, simply because there are so many fraud hypotheses to test.

The best proof of fraud would be a conviction. I'd like to see Blackwell convicted. His obstruction of the Ohio recount was criminal, no?

[Name removed]

23. Deleted message

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

[Febble]

27. Well, read the paper

it's all there.

[tommcintyre]

67. I actually have not only read your paper...

I printed it out, and critiqued/annotated it.

Taking EVERYTHING into consideration (the total "environment" in which this paper exists), I am leaning toward the concept of "red herring", rather than "straw man".

A red herring is defined as:

"Something that draws attention away from the central issue."
http://www.answers.com/topic/red-herring

"...an irrelevant or immaterial... issue."
http://www.legal-definitions.com/Q%20R/red-herring.htm

<word origin:>
"...drag a red herring across the trail to mislead the dogs.
http://www.wordorigins.org/wordorr.htm

------------
Whether it is your intention or not, support/promotion of the rBr "guess" fuels the Bush regime's attempt to create doubt that the exit poll discrepancy points to fraud - as, historically, it does.

Further, contrary to your statement from the post above: "I am not supporting the rBr."; the following statement from page 21 of your paper indicates otherwise:
"The pattern instead is consistent with the E-M hypothesis of “reluctant Bush responders”..."
http://www.geocities.com/lizzielid/WPEpaper.pdf

I think THAT is about the clearest declarative statement I've heard from you so far. ;)

And Febble, you can mince words all you want about the testability of rBr. The fact of the matter is you can only infer the validity of this hypothesis - you can NEVER verify it. Read below for more on this:

Experiments vs. Surveys
EXPERIMENTS
Collection of data from observations sampled from a population that are either treated or controlled by the experimenter. <Inducing the rr to reveal who they actually voted for would've been the best application of this idea - unfortunately, the time has past forevermore for this to be done.>
SURVEYS
Examination of a system already in operation <the data you are working with from E-M> in which the investigator does not have the opportunity to assign individuals to different conditions. <This ALL that can be done now, using inferential statistics.>
-----------
Experiments vs. Surveys
Both are valid forms of analysis, but each
varies in approach and each has their
respective pitfalls and caveats.
- Surveys not as clear-cut as experiments.
- Experiments always preferred when possible.
- Surveys are useful for establishing patterns.
- Experiments must be used to verify patterns.

From:
"The Scientific Method & Basic Statistical Procedures"
http://www.plantbio.ohiou.edu/epb/instruct/quantmet/lectures/pdf/lec6.pdf

Since you can not EVER verify the rBr (since that time has past, and can never be recovered), the very idea is rendered moot.

[TruthIsAll]

10. ALLEGEDLY?

"However, it is true that the "differential response" hypothesis predicts that response rates should actually be lower in Bush strongholds than in Kerry strongholds - because allegedly, Bush voters don't like responding".

THAT SINGLE WORD ALLEGEDLY REFUTES THE WHOLE RBR ARGUMENT.

What is the hypothesis?
That Bush voters don't like responding?
And this accounted for the "differential response"?

Has "differential response" ever been used as an hypothesis in exit polling? Was it used in the Ukraine?

How come, Febble, the Null Hypothesis was not stated as:
This was a clean, fraud-free election.

Isn't that a superior basis for doing real "scientific" analysis?

[organik]

12. MoveOn

The way to get US Count Votes the $ is through MoveOn...they're a cash generating machine. Now, the problem is convincing them to get behind this. Any ideas?

[davidgmills]

13. Pretty hard when they don't believe in election fraud

[Febble]

19. I believe in fraud.

I just don't believe the exit polls have yet demonstrated it.

I think Blackwell's behavour in Ohio was pretty conclusive though.

[TruthIsAll]

15. It's like convincing Al Franken or Bill Clinton or John Kerry...

Where do they want to MoveOn to?
Why the state of denial?

It's downright sinister.

[Febble]

20. If you mean me

I'm convincable. I've spent the months since November 2nd trying to be convinced that the exit polls meant that Kerry won. Why else would a fifty-something British woman be wasting her time trying figure out whether you guys actually elected your president? Especially as you probably didn't mean elect him last time? Why would I be sending data analyes to John Conyers, Cliff Arnebeck, the Kerry Edwards council in Ohio? Writing to Keith Olbermann trying to draw attention to the scandal of voter suppression and possibly worse in Ohio? Trying to figure out whether touch-screen voting in Florida really was biassed for Bush? Check out my papers on the USCV website.

But, not being in a state of denial, I'm not going to lie to myself. If I think an inference is not supported by the data, I'll say so.

But check what I am saying first.

To quote the Cat-in-the-Hat"

"To find a missing something, you must find out where it's not".

I don't think it is in the inferences drawn by USCV in this particular paper. So we can look elsewhere.

I'd start in Ohio. In Columbus. In the strategy Ken Blackwell used to allocate voting machines to urban precincts. And in his illegal obstruction of the recount.

[Name removed]

22. Deleted message

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

[Febble]

24. Numbers tell everything

Exactly. It is numbers I am looking at.

To be even more exact, I am looking at the Within Precinct Error numbers. This tells us how wrong the predictions were for each precinct, after what E-M considered appropriate weighting. And they were massively wrong. By which I mean they massively didn't match the count.

It is impossible that this could have happened "by chance". Which means something very precise also. We can calculate the "sampling error" which depends on the number of people sampled and the proportion of each who said they voted for each candidate. The error here is way way way outside sampling error. So we know the prediction didn't match the count.

But when two things don't match, the numbers don't tell you which was wrong. What you have to look at is correlations.

Here is a testable fraud hypothesis: if the mismatch was due to fraud, and fraud was more common in precincts where electronic voting was used, then you would expect the error to be lower where there were paper ballots. Hypothesis supported. However, there is a problem - paper ballots precincts were small in number and mainly rural. There's a confound. But it's perhaps suggestive. Needs further probing.

Here is a testable rBr hypothesis: if the mismatch was due to Bush voters being less likely than Kerry voters to participate, then you'd expect the mismatch to be greater in precincts where it would be easier for Bush voters to "escape" - large precincts, inexperienced interviewers, precincts where the pollsters were a long way from the precinct. Hypothesis supported. However, there are several problems. One is the WPE itself which needs fixing, and I've fixed it. Another is that if Bush voters were less responsive than Kerry voters, response rates should be lower in high Bush precincts.

So here's another testable hypothesis: if the mismatch is due to reluctant Bush responders, response rates should be lower in high Bush precincts. Hyoothesis not supported. There is no difference in response rates (no significant difference). However, there is a problem here - we do not know the variance in response rates. So we cannot tell whether the reason we failed to support the hypothesis is because there was no effect or because there was insufficient statistical power. It is a question that needs addressing. It is one we should be seeking answers to. But we do not have the data, so the null is still retained. We have not proved that the response rates in high Bush precincts were higher than in high Kerry precincts (which might suggest fraud), but we have not disproved it either.

You are right. Numbers tell the story. If they can be heard over the noise. And it is very noisy data.

[Name removed]

44. Deleted message

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

[Febble]

47. We can't re-analyse the data

because we don't have the data.

One point of my paper was to show that the meagre data points provided in the E-M report are not enough to refute rBr.

But I hope that my contribution to the debate will encourage E-M to reanalyse the data.

I am not working on the opposite side to TIA. S/he seems to think I am.

[Name removed]

48. Deleted message

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

[Febble]

18. Null hypotheses

That statement does not refute the whole rBr argument because it is not true. We do not know whether or not response rates were lower in Bush strongholds because they were not statistically significant from the rates in other precinct categories. This means that there is not enough power, either because the numbers of precincts in the extreme precincts is too low or because the variance is too high, or both. But does E-M report not give us the variance.

How you state the null hypothesis depends on the nature of your hypothesis. If a clean fraud-free election was indexed by values in your variables of interest being equal, the null would be a clean fraud-free election. If a clean fraud-free election was indexed by values that are different in your variables of interest, then your null would be fraud. As you say, you are actually better off with the null being "clean", as it means any effects you find allow you to reject the null. Failure to find an effect does not "prove" the null, only allows it to be retained. .

Statistics is blind. You test not innocence or guilt, but whether things are the same or different. All I have done is to demonstrate that bias (and don't jump on me - bias is fraud neutral) was not significantly different in high Bush precincts. The null is retained. The error was randomly distributed. Interpret that as you like.

If you want to demonstrate fraud, you need to frame a testable hypothesis. One hypothesis I would like to see tested is: was the bias greater in precincts in swing states? Or: was the bias greater in Ohio? The null in both these instances would be "the election was fair". This is the best way round. For your purposes. As you say. It is also what I have been doing so far.

[Chi]

60. Sorry to bother you.

Heya Feeble (waves)
Welcome to DU!

I have a couple of questions for you, if you have the time and patience.
If not...I understand.
Be fore warned, I am blue collar, so you might have to type slower ;-)

My questions are focused on the last 613 respondents in the N. Exit Poll.

Considering that all 613 were Bush voters, and the S.O.P. of Exit pollsters is to adjust the final exit poll to the results, is it reasonable to consider that last group of additions fictitious (a correction <'fudge'>, not an actual scientific sampling)?

If so, did you attempt to filter it out of your data (if that is even possible) ?

If not, could 4.5% of your data possibly being 'fudge', sway your results?

Is there an easy way to re-run your hypothesis and try and filter this data out?

Thanx in Advance

(my apologies if you have responded to this earlier. I made an attempt to read all your posts before I posted, and I didn't see any.)

[Febble]

64. I think people are reading more into my study

than it deserves!

I analysed the data that was given in the Edison-Mitofsky report on the exit polls. They narrowed down the problem to what they call "within precinct error" or WPE for short. The WPE is the discrepancy between the predicted vote count in each precinct (estimated from the responses they get) and the counted vote in that precinct. Obviously if the vote count is wrong, the answer will be wrong. But if their estimate is wrong, the answer will also be wrong. But you can't tell from the WPE which is which.

However, there is another problem with the WPE which is that it exaggerates the discrepancy in precincts in which the support for each candidate is even, and makes it too small in precincts which are highly partisan. This is just an illusion of the numbers however, and it makes any analysis based on the WPE suspect.

This includes Edison-Mitofsky's own results as well as the conclusions (in my opinion) of the USCV report.

My paper suggests that Edison-Mitofsky reanalyse their data using my fix. Then we can see if they still reach the same conclusions. It would be good if they also tested some fraud hypotheses directly. I hope they will.

This doesn't answer your question, but I hope it makes it clearer what my paper was about! A few people seem to have mistaken it for something it was not.

[Febble]

21. See below

But I'm using the word precisely because I am not claiming it as fact. If the allegation is that Bush voters responded less, then the prediction is that response rates will be lower in high Bush precincts.

This is what is tested.

Boy, you guys are paranoid.

And yes, the null hypothesis is that the election was fair. That is the best hypothesis for your purposes, as you state, because you cannot prove a null. You can only disprove the null. We are attempting (yes, me included) to reject that null. We want to demonstrate that fraud occurred. If we can, we are home and hosed.

If we can't, the null, i.e. that the election was fair, is retained. I am demonstrating that in this paper the null is retained. Until we test another fraud hypothesis.

Bring em on.

[TruthIsAll]

25. Now we are paranoid?

Why would you say that?

Because the election was stolen in 2000?

Because the election was stolen in 2002?

Because there is no paper verification of the votes from Republican owned touchscreens?

Because Blackwell refuses to cooperate with Conyers?

Because E-M refuses to release complete precinct data?

Because 43 of 51 states deviated to Bush (odds of 1 in 2 million)?

Because all 22 Eastern Time Zone states deviated to Bush (odds of 1 in 4 million)?

Because the MoE would be exceeded in at least 16 states, all in favor of Bush (odds of 1 in 13.5 trillion)?

Because the Media won't print one line about USCV?

Because we have to debate the RBR?

You are right, Febble.
We are paranoid.

Aren't you?

[Febble]

26. Well before you get paranoid about me

just remember that I am, like you, trying to find out what happened in that election.

Do a bit of googling and you'll find out. Or try the USCV website.

If Blackwell isn't in jail it's not for want of evidence from me.

[davidgmills]

29. Lizzie here's what you need to know about TIA

Most if not all of these analyses are his and he is a hard core math guy with several degrees in Math.

Some people fault his assumptions but no one questions his math skills.


http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=364590&mesg_id=364590

[Febble]

30. Not questioning the math

Questioning the input

[TruthIsAll]

31. What input? Be specific.

.

[Febble]

32. Well they weren't on this post

but I think some of your assumptions regarding noise have been a little odd.

Most discussions about statistics break down over the issue of noise.

[glitch]

33. Noise being a big issue when studying natural systems

but is it the biggest issue when studying something unnatural, like fraud?
Patterns play a bigger role in that case, IMO.

[Febble]

36. Noise often has a pattern

and if you can model the pattern you can filter out the noise.

Fraud may have a pattern that looks like differential non-response (vote switching, ballot stuffing, or ballots spoilage would all have the same signature on the data as differential non-response (rBr if you like).

Margin snipping (perhaps via hacking) might leave a tell tale signature on the WPEs (and more so on my bias index) if the hacker increment to the margin was additive rather than multiplicative. So that's a finger-print we might look for.

Other things to look for, IMO would be greater levels of bias (using my index, not the WPE) in specific places, e.g. swing states; precincts with touch screen machines (although you have to be careful that this is not confounded by other demographic variables like county population); Ohio.

On the other hand, greater levels of bias in precincts where sampling protocol was difficult to adhere to for various reasons would support differential non-response (aka rBr)as a factor contributing to the over-estimate of the Kerry vote.

[glitch]

63. My point is you are trained to look at behavior in a fairly natural state.

Those are the assumptions you are entering into the study with. Very normal thing to do as a scientist, but I think you need to look at this system as being potentially unnatural, possibly manipulated by an external agenda, perhaps fraudulent.

Different beginning assumptions initiate different views and therefore different studies and different results. That's a standard hazard.
And it's why I think you would be wise to try and look at the situation from as many sides as you can, not just the one you are used to.

[Febble]

65. You are absolutely right

And I do.

[TruthIsAll]

34. The issue of noise is a contrived one. Look at my posts.

Again. Be specific. I have made no assumptions about noise. You are putting words in my mouth.

I STRONGLY SUGGEST YOU POINT TO A SPECIFIC POST OF MINE (YOU HAVE MANY TO CHOOSE FROM) AND CRITIQUE THE DATA AND CALCULATIONS.

THIS IS NOT ENGINEERING.

WE ARE NOT STUDYING THE BEHAVIOR OF SUB-ATOMIC PARTICLES.

THIS IS PROBABILITY AND STATISTICS APPIED TO POLLING

42 out of 50 states deviating to Bush is not noise.
16 states deviating to Bush beyond the MoE is not noise.

Consistent National Exit Poll timelines showing Kerry with an average of 50.80% across all the demographics is not noise.

The Law of Large Numbers is what we have going for us.
We need nothing more.

[Febble]

35. So many choices, so little time...

Exactly, it is not engineering.

I think this is at the bottom of my differences with Bruce, who is an engineer.

I am a behavioural scientist, and noise in human behaviour does not behave in the way that it does in engineering.

A behavioural scientis is usually very pleased with an R squared of even 30%. We know we are lucky to model a fraction of the variance in our data.

In my paper I postulate high degrees of variance for response rates. I believe these are justified by the absolute WPE values given in the E-M report. If anything my noise levels are too low. And yet it leaves insufficient power to detect the pattern noted by USCV as "fraudulent".

But we know, as I have said countless times, elsewhere and on this site in the past couple of days, that the overstatement of the Kerry vote was not noise. It was not within the MoE. It was not sampling error. You yourself have computed the probability of the Kerry overstatement being due to chance. I'd cap your estimates - using the WPE alone, I got something like 1 in 9 billion.

It was not due to chance. It was due to something else. Could have been fraud. Could have been differential non-response.

But whatever it had would have had VARIANCE. It is the amount of variance assumed in this mystery factor, this latent variable, that is at the root of the disagreement between USCV and me. And I am fairly confident, because I am not guessing, that my noise estimates are nearer the mark.

Because human error, and also human fraud, are noisy.

[davidgmills]

37. Can you identify the Bush non-responders as being

Bush 2000 voters who were quiet this time after responding the last time, Gore voters who switched and were ashamed to respond to pollsters they responded to last time, or were they new voters?

There were limits on the numbers of Bush 2000 voters.

Kerry won the new voters by a large margin and we had 20 million new voters this time.

So "who are these guys" to use a quote from Butch Cassidy and the Sundance Kid. They must have been all new voters that skewed the new voter numbers in favor of Kerry. Is that what you are saying?

If you are going to have a source for the hypothesis it must come from somewhere.

[Febble]

38. Can you explain?

What do you mean by limits?

And how do you know that Kerry won the new voters by a large margin?

What data are you referring to?

[davidgmills]

39. Ask TIA, he's got the numbers

On new voters.

[davidgmills]

40. Generally though

105 Million voted last time and we are pretty sure about 4% died.

120 Million voted this time meaning approx 20 million new voters.

Exit polls say new voters went to Kerry by a substantial margin and Bush and Gore voters stayed about the same so Kerry won unless:

(1)Bush voters who responded in 2000 didn't respond this time,
(2)Gore voters who responded last time and switched didn't respond this time, or
(3)Polls were really wrong about Kerry's margin of victory for new voters because new Bush voters refused to respond.

So which hypothesis is it?

[Febble]

41. I don't know

No-one does.

On the basis of the E-M report I'd say the evidence supports differential non-response as a factor. With mean response rates at only just over 50% it is certainly possible. But I'd want to see the analyses re-run with my variable before I was completely convinced.

[davidgmills]

46. rBr is counter-intuitive

By the way, welcome to DU. We only mean to challenge not insult.

Since you are British, you can not possibly understand how counter-intuitive the rBr hypothesis is to most of us who have followed American politics for years.

If the hypothesis were the "reluctant Kerry responders," most of us would say "OK, I can see that."

Even Rick Brady, whom I have emailed numerous times and I know you have as well, concedes that this hypothesis is counter-intuitive. It is as if Mitofsky hit us over the head with a brick.

Democrats, by in large, are urban and except for the Democratic "elitists," (as they love to condescendingly refer to those of us who are educated), tend to be poor, tend to be minorities, tend to be products of poorly funded urban schools, tend to have to vote in precincts that are less than desirable, tend to be strapped for time, tend to be intimidated by people of higher financial and educational status (pollsters), while the reverse is true for Republicans. What you describe as despicable in Ohio has been a truism for most of us most of our lives. We are used to this kind of treatment of our disenfranchised.

I am fifty four and have been an attorney for 27 years. I represent the disenfranchised of our country far more often than I do those of the middle class or higher. I know these people. They don't rush pollsters like linemen rush a quarterback dropping back to pass. They just don't do it.

I have followed politics in America closely for 37 years or so. Many on this board are like me are incredulous of this hypothesis. We understand that poor response can mean poor polling. But we seriously doubt the poor response was by Republicans. It is so antithetical to what we witness every day.

So please forgive us, if we take you on.

[Febble]

50. Thanks for the welcome

There has been a bit of a misunderstanding over my position on "rBr" here. USCV refuted it using a bit of analysis that I don't think stands up to scrutiny (although at first I found it at least an intriguing argument, which was why I pursued it). So my conclusion was that "rBr" remained unrefuted, not that it was true.

We do not know why in so many precincts, the the polls over-estimated Kerry's share of the vote. It could have been fraud, it could have been sampling bias.

What we do know, in a statistical sense, is that the over-estimate was greatest where polling protocol was likely to be poorest (large precincts, inexperienced interviewers, etc). This suggests that poor polling protocol played a part.

How could it do this? One theory is that while both sets of voters were reluctant (50ish percent is a poor response rate) Bush voters, in any given precinct, tended to be more reluctant. It could well be that in poor urban areas, sampling rates for both groups was lower. But this would not affect the hypothesis, which just requires that for a given neighbourhood, Bush voters would be slightly more reluctant to participate than Kerry voters.

It may be wrong. There may be some quite different sampling reason. It may be that Bush voters tended to vote early, and Kerry voters late, and the interviewer had got the hang of things by then. Or that Kerry voters tended to vote in the middle of the day when the crowds were quieter and there were fewer misses. I'm not saying any of these are true. But if there is a significant correlation between the extent of the over-estimate of Kerry's share of the vote and the extent to which good sampling protocol was likely to be at risk, then it is strong evidence that polling error was at least part of the problem.

However, I think the jury is out on this because of the problem with the measure that is detailed in my paper. I think that if the analyses were re-run using my measure, things might become clearer. Including whether or not the over-estimates were correlated with factors that might suggest fraud.

[mgr]

42. Be careful about definitions

David:

There may be a difference between a new voter, and an intermittent voter, it depends on how the question is phrased. If one is a registered Republican that did not participate in 2000, but did in 1996 and 1992, one is not a new voter. A new voter may be one that registered for this election with the intent of voting for the first time.

Mike

[davidgmills]

49. Woopie!

And that adds how much to the argument?

[mgr]

69. In this way

David:

Is your estimate of new voters independent of NEP's exit poll? If not, then you have to bracket it by an MOE of 6.5% minimum. Move this downward by the MOE (since this is counter the direction of the bias), and assume static B/G voter support (the active voters from 2000), and the infill between will be casual voters.

The thing that may affect this further, is that a casual voter may intend to have voted in 2000 but did not, but believes that they had. This is how I would characterize the bandwagon effect, though I have issues with this explanation based upon the studies cited that I have reviewed so far.

Mike

[TruthIsAll]

43. Kerry's New Voters? Febble, you need to educate yourself.

I have asked you to read my posts.
What are you waiting for?

They leave no stone unturned.

All numbers.
All facts.
No conjecture.
No BS.

Welcome to DU.
You will learn a lot here.


http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x357758

[mgr]

45. Differential response affects your analysis as well

Febble does not need to respond, I have already commented that if NEP is accurate in its characterization that it improperly or poorly trained their sample takers, and the bias they report exceeds the MOE, you cannot apply the MOE that was part of the original sampling design, but a larger one. When one puts that error term around your means, much of the problem falls away.

The problem that you do not seem to appreciate, is that she provides a correction term that may allow you to obtain what you seek. Why don't you analyze the LA Times Exit Poll, see if the same problems exist? It has been completely overlooked in this conspiracy laden popular vote was stolen argument of yours.

Mike

[Febble]

51. Thanks.

I am not disputing TIAs math.

But I think the evidence that poor polling protocol was a significant predictor of the extent of the mismatch between pool and count is strong evidence that it played a part.

The difference between TIAs assumptions and mine are simply that I am more prepared to accept the possibility that the polling went wrong. Collecting data is difficult, and at bottom, the polls are a sample and the count isn't - or shouldn't be. Any sample is prone to both sampling error and sampling bias. Vote counts, sadly, may also be prone to "sampling bias" aka fraud.

But whereas we know that the polls were a sample, and prone to all the ills that samples are heir to, we don't know that the vote count was.

I understand that some may feel that vote count corruption was so likely, and so likely to be vast, that it competes seriously with the possibility that the polls were simply wrong. But the idea that polls are unlikely to be wrong is simply a myth. Sampling error is simply one of the errors that polls can make. Sampling bias, for a whole host of reasons, is another, and the lower the response rate, the more likely - inevitable - it is. And retrospective questions are notorious for getting impossible answers. Polls in Britain conducted a couple of years after an election, tend to show that most people voted for the currently most popular party, not the party that actually won. And the results are well outside the "margin of error".

[TruthIsAll]

52. Bruce O'Dell at USCV has already refuted your model. RbR is dead

..in the water.

USCV deals directly with your model assumptions. I will not delve into that.

I have shown that E-M refutes the RBR conjecture with their own data in the National Exit Poll AND YOU HAVE NEVER ADDRESSED THIS CONTRADICTION.

HOW VOTED 2000:
13660 Respondents: BUSH 43%/GORE 37% (IMPOSSIBLE, BUSH WINS 51-48))
13047 Respondents: BUSH 41%/GORE 39% (IMPLAUSIBLE, KERRY WINS 51-48)

PLAUSIBLE: BUSH 39%/GORE 40% (LIKELY, KERRY WINS 52-47)

[Febble]

55. Bruce is entitled to his view.

And I have every respect for him.

We have had a good and co-operative working relationship which I hope will continue.

At the moment he believes he has refuted my case. I beg to differ. I think he has underestimated the variance in differential non-response, as measured by the absolute WPEs given in the E-M report, and the ability for high leverage points to impact more on the mean when Ns are low than when they are high. I also think he has underestimated the variance in the response rates.

And I have addressed your other point on another thread.

[mgr]

54. I think that many on this site forget the Washington State recount

One of the early arguments was that election results on the west coast (where I am from) were not tampered with. The gubernatorial race was far closer than exit polls predicted, and served as the best laboratory for the fact that the red shift was an artifact of polling, not the election conducted in Washington. There is of course the suit addressing the discrepancies with the DREs, but it is limited to one county, and not the largest in population. Everyone also assumes that California's was clean due to Kevin Shelley, and Oregon's has sufficient safeguards, and meshed perfectly with a telephone survey exit poll. This is nearly 10-15% of the election population.

As more and more states can be argued to have run clean elections (not perfect, but clean), the more the argument fails. What we are left with is having to look at the battleground states, and determine if their results are the outcome of vote tampering. Unfortunately, TIA's work gets in the way of this.

Mike

[glitch]

61. I would never assume the elections on the west coast were clean.

Some serious padding was possible and IMO likely on the west coast as well.

[mgr]

68. How so?

The point with Washington is that the hand recount indicated that exit poll overestimated democratic support, if there had been padding of this via electronic means this would have been evident. Oregon was a telephone survey of mailed ballots that have several layers of security unique to their system. California did not allow paperless DREs, and being stongly democrat through out is unlikely to permit any shenanigans.

One may argue that compilers were hacked, but if not evident in Washington's recount, then it may be true for other states, unless Rove was prescient that a recount would occur prior to the election taking place?

Hacking compilers implies your opponent is smart, and able to change--the antithesis of conservatism. If your opponent were smart why do we have the goods on Blackwell, is he an exception, or the rule? I think a phase shift in thinking is in order here, people are stupid, if it worked in the past, they will continue to do it in the future. Look for ballot stuffing and voter suppression first, go to the compiler argument later.

Mike

[Name removed]

53. Deleted message

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

[Name removed]

56. Deleted message

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

[TruthIsAll]

58. Wrong, Febble. I have considered your work and reject RBR, as does USCV.

There is nothing further to debate.

[Febble]

59. Fair enough.

[mgr]

57. I am sorry but

As I noted to another poster, I take the professional opinions and analyses of behaviorial scientists over those of an engineer in this matter. The methodologies associated with polling are the bread and butter of their profession. As to irrelevant and annoying...it must be to you since Febble's and my world views on the matter are far more consonant than with yours, or do you recall your quibble with my point that your methods and approach to the matter are not in line with USCV's, or that I think our effort better focused on battleground states, or that do we really need to argue that exit polls are right to support our contention that fraud occurred? Stop making me into your strawman naysayer, since I have strongly argued that fraud occurred. Soon, you may have the opportunity to carry my lunch.

Mike

[glitch]

62. Behavioral scientists look at behavior, so their results depend on what

they are trained to look for.

I want to see the results from the people who are experienced at determining fraud from patterns in numbers.

Forensic accountants and data experts are who I want to hear from.

[Febble]

66. Yup, they are the ones we want.

but I hope I have helped by giving them an uncontaminated variable to work with.