What would it take to get access to the raw exit poll data?

[merwin]

That is the ONLY way this can be finalized, once and for all. We know demographics already, and if we have the raw data then we can (fairly easily) weight them the way they are supposed to be weighted, with full documentation as to why.

I think that we should shift from 'Here's what I think happened' to 'If you are so confident that you are correct, show us the raw data'. And if they refuse, schedule protests outside of Mitofsky's residence/workplace on a regular basis, since he's the one claiming 'trade secret'.

[Wilms]

1. John Conyer's asked for it. I don't know if he heard back.

The Dems taking both Houses next time around would be a way to stretch the currently short arm of the law.

[davidgmills]

2. A class action suit by 20 A G's

[merwin]

6. Hm... could the people start a class action lawsuit against Mitofsky for

failing to provide the services that we as citizens paid him to provide?

Hasn't he said already that the exit polls were wrong?

[davidgmills]

26. NO

No privity of contract.

But Ag's might be able to demonstrate enough of a state interest to get it.

[Andy_Stephenson]

3. A revolution

and violent overthrow of the government?

:shrug:


(not inciting here...just a rethorical question)

[merwin]

5. It's so sad that people can't see through the web of deceit. And to top it

off, most people will avoid confrontation or protesting or even acknowledgement until it directly affects them.

[garybeck]

4. a subpeona

same as cracking open the voting machines. need a pending lawsuit, a lawyer who requests it, and a judge to approve it.

[rigel99]

12. YES LAWSUITS

beck has it right again. as usual, Beck is right again!

folks.. watch out next week for some really fun lawsuit action....

[skids]

7. In addition to the already released raw data...

...what is the important thing you consider to be missing?

I mean, some things are missing, granted. The results in discarded outlier precincts, the sampling rate, etc. We are slowly figuring out the real-world identity of the precincts (we can already figure out what county most precincts are in due to a sort ordering) but what among those items do you think is critical to taking the analysis beyond what we can do with the released raw data?

[kiwi_expat]

8. The data is available at www.icpsr.umich.edu/org/announce.html#nep

(See "Ohio Exit Poll Raw Data" thread.) It is surprising that most DUers seem not to be aware of that.

The remaining hurdle is getting the exit poll precinct names from all the BoEs - 88 BoEs in Ohio, alone. (We can then immediately match up the exit poll precinct names with the data.) Maybe there is a way of getting the exit poll precinct names from one office in each state. But I don't think the SoSs' offices will be very helpful in states like Ohio and Florida.

I assume it would be necessary to know the identity of the exit poll precincts, in order to weight their data for a test - maybe it wouldn't. It would certainly be necessary to know the identity of the exit poll precincts, in order to compare the NEP precinct data with the official precinct vote counts.

[FreepFryer]

9. That is the adjusted, not raw, unweighted data. Not useful.

Not usable for anything other than forensics (not statistical analysis). Also not identified at the precinct-level.

[kiwi_expat]

10. minvis has downloaded and deciphered that data

It is definitely precinct-level data. As for weighting, it is my understanding that any weighting is clearly identified. I assume that the weighting at this level is the missing-person/no-response weighting that Mitofsky says he applies at precinct level.

I am pretty sure that it is not the weighting that he applies to each precinct, according to his model, to get the aggregate totals.

[FreepFryer]

11. As I said, it's not _identified_ at the precinct level. (n/t)

Therefore any analysis is supposition and number-matching (forensics), not true statistical analysis.

Just the way the NEP likes it.

[kiwi_expat]

14. It is very easy to identify the NEP precincts...

once a county/state set of exit poll precinct names is obtained. We know the sequence involved in the matching. (With no help from NEP, other than their laziness in assigning code numbers.)

From that point, "true statistical analysis" can be done.

[FreepFryer]

15. Right. And that point HAS NOT ARRIVED. (n/t)

[kiwi_expat]

16. At least we now agree....

that the file includes the "raw, unweighted data" (see 9). :-)

[FreepFryer]

17. No, we do not. Repeating the untrue sadly does not make it true.

Please see all my posts.

no one would be happier than I to be able to admit we had the actual raw, unweighted exit poll data.

We do not. We have the adjusted, weighted data ONLY, and a 'leaked' set of data whose origins and validity are not beyond reproach. The identification of the weighting present in the NEP data at that site (which I found about 2-3 weeks back) is inconclusive towards a determination of the direction and cause of the weighting. It's as if you have a total sales tax amount for a purchase and are trying to determine how many items you bought. Unless you know ALL the prices, you cannot crack the statistical variance of the cause. You can perform forensics, based on previous purchases and 'average retail prices', as TIA has done with aplomb and great endurance, but you cannot make any assertions that themselves do not comprise an appreciable MOE.

TIA's posts have been brilliant, and edgy, and provocative, and accurate, and factual. That's not the issue. The fact is, you are mischaracterizing and oversimplifying the state of this data by calling it 'raw and unweighted'. In addition, the comparison you propose is indeed forensics, not statistical analysis, unless the data can be independently verified (an entirely different process than 'matching' totals by sequence to precincts).

This has been discussed here since before the election, ad nauseam. I'll leave it to others to explain it again... but please, try to take one, intelligent step at a time and not race for an imaginary finish line, 10 feet beyond the cliff.

[skids]

27. Okay, hold on... just what data set are you guys working with?

...because the data I have has one line per respondent, with the exact answers that each respondent gave on the survey.

That, in all respects, is "raw" data. The only thing missing is the surveys from discarded outlier precincts, the exact real-world mapping, and, unless it is available someplace else, the sampling rate/completion rate/refusal rate.

[FreepFryer]

28. The data set mentioned above, in this thread. (see post #8).

This includes the respondents added to provide 'weighting' of each state's results, as TIA mentioned.

In addition, there is no proof that the data in these results is the actual responses of the actual respondents.

It's packaged data, weighted to support the NEP's conclusions, in no way verifiable. There may be correlations drawn about fraud, but there is no way to prove causation, under these circumstances.

In the opinion of many, this (published) data is fraudulent... and so cannot be used to prove fraud. And it is impossible to move from 'opinion' to 'fact' without independent verification of the RAW data, which has never been made available.

[skids]

29. OK, to clear something up.

There is no proof that the "added respondents" were a biased selection -- the weights were readjusted, so the added respondents could have been anything. The final weighting problem is much more likely just that -- a final weighting problem, not the addition of bogus survey responses, the addition of bogus weights.

Secondly, if NEP was going to create fraudulent survey entries, they would have done a better job at it than they did, because some disturbing patterns are evident simply by looking at that data. That data does more to raise questions than calm things down.

You can never have "proof" that the results are real. Even if they collected and turned over the paper documents themselves, that still would not be "proof" since they could be forged.

As an aside, "proof" is not required. Not here and rarely in a court of law. That's why we have juries, to weigh the evidence. Exit polls are not proof, no, but they are corroborating evidence.

All said, it is true that it would have been nicer to have the data in separate files corresponding to different early releases, and the early release weighting. However to say that the released data is not "raw" is not accurate. It would be more accurate to say what you are looking for is the set of pre-7pm respondents. (There is a good chance that order of the respondent ID's in the file is significant, by the way, FWIW)

[FreepFryer]

30. If proof is not required, is yours a 'faith-based' case?

Other than that somewhat odd assertion, we are in agreement.

The exit polls are not proof. They can correlate, and thus imply causation vis-a-vis fraud, not prove it. The data presented is likely flawed, if not outright fraudulent.

However, in the legal system of the United States, proof of guilt is indeed requisite - at least ostensibly. And that's the litmus test in a politically divided nation where juries can be selected to achieve any desired result.

[skids]

31. No, it isn't.

Talk to L. Coyote about what "proof" means. He's a lawyer. He can set you straight.

The legal definition of "proof" as in "innocent until proven guilty" is that a jury decides. Proof in that sentence means "decision of the jury." This is why, in murder cases, they have to tack on "beyond a reasonable doubt." If the kind of "proof" that you carry on about was required in all court cases, then that language would not be needed. Also, if a simple decision by a jury was not legally considered "proof," then the newspapers would have to keep saying "alleged" even after the verdict, whereas they are legally immune and allowed to stop doing so even if the verdict eventually gets overturned.

Geesh. Mathemeticians. How do you manage to cope with the rest of reality? :-)

[FreepFryer]

32. I'm pretty clear on 'proof'. In this case, we ain't got it.

We can believe all we want, we can 'make a case', but unless we do all our homework, we'll lose in a stacked game.

That's all I'm saying... we need to win this one, not just put up a valiant, ultimately futile, struggle to assert what we believe to be right.

And most of all, we need to avoid declarations of proof when there are none... each time, we feed the Rove-o-tron and weaken our valid case with misfires.

This battle is unlike any other - certainly unlike any other that has ever seen the inside of a courtroom. This one is for all the marbles, and the rules aren't the same as the court of public opinion.

Bush must, and will, resign. But it won't be because of a good case in court. It'll be because we have his nards in a ringer.

[sacxtra]

34. YOU WANT PROOF?

GIVE ME LEGAL ACCESS TO ANY PART OF THE ELECTRONIC VOTE SYSTEM.

I WILL TRASH THE FUCKER.

[rigel99]

13. Where can I find Exit Poll Data for Georgia

need it for some of my auditing and analysis

[kiwi_expat]

19. I suggest sending a PM to minvis.

[kiwi_expat]

18. Blue22 has put the (Univ. Mich.) NEP Ohio precinct data in Excel format.

He has kindly offered to e-mail the spreadsheet to anyone who PMs him their e-mail address.

See the data for yourself!

[FreepFryer]

20. *yawn* hundreds of us have the data already. Understand it before you

represent it for what it is not.

[kiwi_expat]

21. The data contains....

"Number in precinct, number in sample, Sample weight, number males, number females, MaleWgt, FemaleWgt"...(etc. for all categories)

What more do you want??

* * *

After posting the above, I see that you have greatly expanded your comment in 17.

Fine. You statisticians wait until you get the perfect set of data out of Mitofsky. In the meantime, we will be busy identifying precincts in Ohio for citizen recounts to PROVE fraud, if we can.

[FreepFryer]

22. It's not what I want, it's what's required for a real analysis.

I only hope you're not trying to spread disinfo on purpose... it's clear how badly you want this to be so.

[kiwi_expat]

24. I am glad you decided to edit your remark.

Please see my edited version of 21, above.

[FreepFryer]

25. We just reached the point of diminished return. (n/t)

You are of course welcome to continue your efforts.

I implore you however to exercise restraint before you assert fraud based on exit polls. TIA's ONLY regrettable assertion was 'the smoking gun'.

[kiwi_expat]

35. No worries. I would never "assert fraud based on exit polls".

This is all rather amusing. I have piped up, making that same point, on some of TIA's threads.

I do, however, hope that the exit polls will be useful at the precinct level (despite the high MOE) to help identify a few precincts to select for citizens' (full-)manual recounts. The manual recounts could PROVE fraud occurred. Statistics and equations can not.

[FreepFryer]

23. Whomever you think 'we' are, you're missing the boat.

Your arrogant dismissal of the requirements of true statistical analysis aside, you cannot prove fraud with exit polls.

You CANNOT. You may indicate its possibility, but until you actually verify the data, you are building a strawman.

Of course, this too has been discussed again and again.

What we should be doing is demanding the data thru legal means (which some of 'US' are doing), while performing forensics (which some of 'YOU' apparently are doing)... not making blanket assertions and discounting the observations of those only seeking to clarify the issues under discussion.

And, don't forget the 'other' us... the 'us' busy trying to turn just one informant over who knows what really happened in Ohio, at NEP, in D.C., and, for that matter, in Iraq and Ukraine.

[sacxtra]

33. RAW DATA. Let us see with a hex editor. No? How about FAKE DATA?

How about I replace your ASIC with MY ASIC?

How about my MEMORY CARD with YOURS?

How about My Logic with yours?

Howa about find the geographic location of ALL the exit data, then go to the DMZ and CUT the wires?

How about just wiping out a few power polls in the local area?
Whoops, "my car crashed."

I ain't suggesting you do these.

Give me legal ACCESS to ANY part of the electronic vote system and I WILL FSCK IT UP.