Mathematical Proof: TIA is Wrong (Part 1)

[Nederland]

TIA claims that exit polls are proof of fraud. In this post I will conclusively demonstrate why this is not the case using a simple example. Applying the KISS method, let's start with a clean election with no fraud. Two candidates (C1 and C2) compete against each other and the resultant voting pattern looks like this:

Precinct ___ C1 ___ C2
----------- ---- . ----
Precinct #1 6500 _ 3500
Precinct #2 6000 _ 2000
Precinct #3 6500 _ 3500
Precinct #4 6000 _ 6000
Precinct #5 5500 _ 4500
Precinct #6 5500 _ 4500

Totals ___ 36000__24000
Percent ____ 60% __ 40%


Remember, for right now our example contains no fraud. The voting pattern above accurately reflects what actually happened--no funny business. Now let's say that an exit poll firm chooses to sample voters as they leave a precinct. The firm chooses precinct #4 and samples every tenth person. Let's say they get the following results, with E1 and E2 being the raw exit poll numbers:

Precinct ___ C1 ___ C2 ___ E1 __ E2
----------- ---- . ----
Precinct #1 6500 _ 3500
Precinct #2 6000 _ 2000
Precinct #3 6500 _ 3500
Precinct #4 6000 _ 6000 _ 592 _ 608
Precinct #5 5500 _ 4500
Precinct #6 5500 _ 4500

Totals ___ 36000__24000__ 592 _ 608
Percent ____ 60% __ 40%__49.4% 50.6%

Now if you look at precinct #4 by itself, the exit polling firm did pretty well. Their sample predicted a 50.6% win for C2, which is within 0.6% of the actual results. Not too bad and well within the margin of error. However, clearly there is a problem. Their results for the entire election are not even close. In fact they are well outside the margin of error for the size of their sample.

What went wrong here? The answer is obvious. They chose their sample precinct poorly. Now many of you might be saying that I cheated and simply choose precinct #4 to prove my point. Look again. There isn't a single precinct that the polling firm could have chosen to get results that would be accurate for the entire election. (It is true that if you happen to select multiple precincts and add their results together you can end up with an accurate exit poll result. I will discuss this further in Part II).

The irrefutable point being made here is that TIA is making one big assumption in the MOE calculation. That assumption is that the samples in the exit poll are representative of the entire voting population. As I have clearly demonstrated in this post, it is perfectly possibly to have a clean election, sample a portion of the voters accurately, and still end up with a result that is well outside the margin of error.

To sum up, this is the lesson of Part I:

Regardless of your sample size, if your sample is not representative of the total population, exit poll results can be outside the MOE.

I will discuss the logical effects of this in Part II.

[Fly by night]

1. Until Mitofsky releases the precinct locations that were exit polled, ...

... we won't know how representative they were. However, since Mitofsky has been in the exit poll business (and has done a credible job until recently), I would expect he knows a thing or two about picking representative precincts.

[Nederland]

4. Yes

I agree. And if we see those numbers released and things look funky I'll be the first to scream foul. Until then, I simply object to TIA's assertion that this is a done deal and fraud has already been proved using merely what we have today.

[Gman]

2. While I haven't put a lot of weight in TIA's figures

by far he's not the only person, including some Phd mathmaticians that have claimed the likelihood of fair elections in electronic voting states was far outside the margin of error. (hope I phrased that correctly...) In other words the results in electronic voting states far exceeded the margin of error, according to many others besides TIA.

[blue4barb]

3. Wasn't Mitofsky's exit poll data used in the Ukraine

which resulted in a second election, overturning the original results? It's hard to understand why he could be so right in the Ukraine and many other elections, and so wrong in November.

This does not compute.

[Nederland]

5. The difference

...is that in Ukraine the exit polls revealed differences between exit poll number and official vote numbers at the precinct level. Those discrepancies made it possible for investigators to know exactly where to look for fraud and compile huge amounts of additional physical evidence of fraud. Until Mitofsky releases more data, we are not in a position to do that.

[skids]

6. Wrong.

Precincts were chosen correctly.

After the election, Mitofski fed the final results into the precinct weighting formula and they were only one half point off from the correct results. In Bush's favor.

If you want to debate the exit polls, read the NEP report and USCountVotes criticism of that report. Until you do, you have nothing constructive to add.

[papau]

8. You beat me to posting - :-) TIA is spot on correct - while Nederland

is correct about the importance of the selection, he is trying to find a reason TIA might be wrong - and is not proving TIA wrong.

And this reason "Precincts were chosen correctly?" question has been reviewed - and despite the media trying to pretend there was no election theft - it is looking like it will go down in the history books as the second stolen Bush "election"

[Nederland]

17. TIA is not correct

TIA has always asserted that you can calculate the MOE simply by taking the total exit sample size and plugging it into Excel. As I have conclusively demonstrated in this thread, that is incorrect. You must take into account precinct level demographics and weight your samples accordingly. TIA's claim that weighting the samples is not only unnecessary, but evidence that Mitofsky is fudging his results:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x341540

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x339246

[papau]

35. I know we all agree - the words "total exit sample size" mean - I thought

that we were talking the stratified sample - and indeed the excell programs tosses out the right MOE.

And his independent stratified state samples indeed can be treated as independant variables - making his multiplications valid.

Or is TIA doing a multiplication that incorrect? If so I'm sorry I have missed the error - as I assumed he was just using (1` - p1) times (1 - p2) times etc. , and p1 times p2 times pe times etc

What prevents adding the state strat sample results to get a multi-state result?

[Nederland]

44. Nope

Look here:

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x339246#339648

TIA claims that the MOE is 1.0%, the number achieved by plugging the raw sample size (13,047) into Excel. Its ridiculously simplistic and dead wrong.

[papau]

52. raw sample is not after stratification model rules application?

As in we have enough females for this 60 minute time interval so I can not ask any more females at this precinct to respond for the next few minutes,or alternatively and more likely I take a random sample of the females that have responded where the sample size obtained fits the stratification requirement.

In the latter case it is obvious that simply adding those sample results "raw numbers" does indeed work to get an MOE calculation input number, and indeed it works for the former case.I really do not believe that the 13000 response count is pre being put through the strat model.

And I am just not seeing dependent variables.

And "Since the poll consisted of "raw" data (even though the sample was randomly selected - see the notes) it needed to be "weighted" to match the recorded vote - along with adding 613 new voters to the original sample, bringing the final total to 13,660 respondents. And thus we now have the official National Exit Poll." is truth.

Indeed we would all agree that the standard procedure of changing the strat model at the end of the day to fit the "recorded actual vote" is a ART and not a science, where the goal is just number matching before we try to find out sub-group trends from the prior election.That matching does indeed come up with a best fit at 37/37/26 D/R/I BUT THAT DOES NOT MEAN 37/37/26 is truth. Indeed that is the point - Either fraud - or a very bad initial strat model - can explain the numbers.

And the strat model Hispanics numbers pre national final fit does indeed match other Hispanic org strat model results. The final fit strat model does not match other org exit poll results.

Perhaps party ID has indeed changed from 39/35/26 - and perhaps Hispanics really love Bush in battle ground states but not in other states - As in 2000 where we had to believe 20000 holocaust survivors voted against the Jewish VP, and that black turn out was low, while absentee ballots went big for Bush, and 10% of black vote could not be read by optical readers vs 2% of white vote. -in 2004 we need to believe there were major shifts - because Bush is a wartime president.The evidence I am seeing makes me believe fraud is the more likely answer.

Indeed the courts do allow stat evidence - as the idiots in Ohio will find in their stupid harassment suit of the BBV protestors/recount requestors.

But You are correct in that one cannot "prove" anything that can be explained in 2 different ways. But why are we not getting access to the computer programs - why are there election day mods being done - and post election removed - all with no record other than "I saw the tech with the machines"?

I think TIA has a solid case - and I see no math error that destroys the case.

[Nederland]

58. Question

You said:

Indeed we would all agree that the standard procedure of changing the strat model at the end of the day to fit the "recorded actual vote" is a ART and not a science, where the goal is just number matching before we try to find out sub-group trends from the prior election.That matching does indeed come up with a best fit at 37/37/26 D/R/I BUT THAT DOES NOT MEAN 37/37/26 is truth. Indeed that is the point - Either fraud - or a very bad initial strat model - can explain the numbers.

Would you agree that when turnout increases dramatically prior election data is less valuable and referencing recorded vote totals inevitable?

[papau]

62. Yes to both - but we all have our "reasonableness" tests

and the independent Hispanic exit polling is a point that will not go away.

This appears - to me - to me fraud - rather than a massive get out the GOP vote success.

The apparent fact that it was at the margins - and only in a few "battleground" states - using different methods from Kerry vote equals Bush vote on some number of ballots - as testified to - to central vote tabulators having internet access to original vote machine data meaning they can change all data so an audit discovers nothing, to a dozen other (actually 40,000 other) incidents that are testified to - does make it hard to chase.

We need the exit poll model - and data - peer reviewed.

[Nederland]

63. Agreed

And perhaps I'm not being clear. My big beef with TIA is the assertion that we have sufficient data to prove our case. I agree with you, until we have full access to data and transparent processes we have only hunches, not proof.

[Nederland]

12. Fair Enough

I sense that you aren't disagreeing with my main point in this thread, namely, that choosing precincts correctly is the key. I sense we will have plenty to argue about in Part II.

[BlueEyedSon]

7. 2 refutations.

1. The law of large (or at least larger) numbers. No one does exit polling - with the intention of getting useful results - and then polls ONE precinct, or a bunch of similar precincts. Give the exit polling guys some credit.

2. Historical good correlation of exit polls and election results.

[Nederland]

10. Correct

No one polls one precinct. However, the percentage of precincts polled in my example far exceeds the percentage polled in the 2004 election. I have therefore demonstrated that it is perfectly possible for exit polls to fall way outside the simplistic MOE calculation that TIA asserts without violating any of the laws of statistics. I haven't proved that is what happened, I have merely proved that it is possible. For now, the only question I have for you is whether or not you believe my final assertion, namely:

Regardless of your sample size, if your sample is not representative of the total population, exit poll results can be outside the MOE.

If you agree with this great--we can argue about the implications in Part II. If not, please tell me why because you haven't so far.

[papau]

54. We all agree that "not representative of the total population" is an alt

explanation.

But why is that a "proof" that TIA is wrong.

[johnaries]

9. But they would have to CONSISTENTLY pick poorly. That's why you

always choose your samples RANDOMLY! The laws of randomness dictate that as long as your sample size is large enough and random enough the results will always form a Bell Curve. Once the curve has been determined by the sample size, there is a 95% Confidence Level that actual results will fall within the upper and lower limits if the limit determination was done correctly.

For what you propose to have occurred, they would have had to use a non-random method and, basically, DELIBERATELY picked the wrong precincts.

Sorry, I don't think your conclusions hold water.

[Nederland]

11. Only one question for you

There is only one point I'm trying to make in this thread:

Regardless of your sample size, if your sample is not representative of the total population, exit poll results can be outside the MOE.

If you agree with this great--we can argue about the implications in Part II. If not, please tell me why because you haven't so far.

[abbiehoff]

14. But there was no sampling error.

TIA is right in assuming that there is no error due to the sampling. Mitofsky and his critics have all agreed that the sample itself was not the problem.

"The Edison/Mitofsky report acknowledges widespread discrepancies between their exit polls and official counts, and admits that the differences were far greater than can be explained by sampling error."

excerpted from http://uscountvotes.org/ucvAnalysis/US/USCountVotes_Re_Mitofsky-Edison.pdf

The sample itself is a non-issue.

[Nederland]

20. You cannot prove that

You assert that there was no sampling error. As I will demonstrate irrefutably in the next part, it is impossible to prove that.

You can argue with me about that later. For now, I think we agree on the principle assertion of this thread.

[abbiehoff]

29. I'm not asserting anything.

Mitofsky and uscountvotes and nine PhDs have said the sample was good. Who are you to say that it wasn't?

[Nederland]

31. Irrelevant

Who I am is irrelevant. All that matters is the logic of my arguments. If my logic is wrong, I'm wrong.

Who I am is irrelevant. That is a fundamental truth of all science.

[abbiehoff]

33. The pollsters aren't just throwing darts.

They know the demographics of the precincts they select and they weigh accordingly based on the demographics of voters in general. No one claims that exit polling is an exact science, but it is a science nonetheless. Your original premise is not sustainable. There is a great deal of study that goes in to which precincts are selected, and if a given precinct is not representative of other precincts, the results are weighted accordingly.

[Nederland]

38. Wrong

They know the demographics of the precincts they select...

Pray tell, where do they get this data?

And if you suspect that I'm setting you up, you are correct :)

[abbiehoff]

39. You obviously know much more about it than anyone else.

You tell me how they choose the precincts.

[Nederland]

42. They are chosen

...by looking at past election results. When turnout increases by as much as it did in 2004, that means that the margin of error increases significantly.

[abbiehoff]

43. I believe they also look at voter registrations

and census data.

[Nederland]

45. Correct

Which adds even more error because not all register voters vote and certainly not all citizens vote.

[abbiehoff]

46. Yes, but there are such things as trends.

In fact, almost magically, precincts in my county that had many more voters in November than ever in the past absolutely maintained exactly the same Democratic vs. Republican percentages. Imagine that.

[johnaries]

15. I agree that if you "cherry-pick" sample data instead of using random

data you are correct. HOWEVER, if your data sampling is truly random then the odds of the results falling outside the MOE is extremely small if your sampling size is large enough. One of the first things we do in Six Sigm analysis is to check our data for randomness before we do anything else with the data. For so many results to fall so far outside the MOE, the odds are astronomical, so as to be negligible.

For your assertion to be true, it would have to have been a deliberate effort.

[Nederland]

19. Not necessarily

For so many results to fall so far outside the MOE, the odds are astronomical, so as to be negligible.

Not necessarily. The exit poll was not far outside the MOE. In order for the exit poll to be inaccurate all that would have to happen would be for the samples to to misrepresent the larger population by a few percentage points. Not "astronomical" at all.

[johnaries]

23. Yes, astronomical. Even for a few percentage points, when it

happens to multiple samplings. The odds against one sampling to be outside the MOE by just a few percentages isn't really that bad. But for multiple samplings the odds don't just double or triple, they rise exponentially. They can get "astronomical" much more quickly than most people imagine.

[Nederland]

26. Prove it

Calculate the odds show me your numbers, just like TIA would. Perhaps I'm wrong.

[johnaries]

34. I'm too lazy, but you might want to check this out:

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=109&topic_id=16137

[L. Coyote]

73. If the methods are flawed, the flaw will impact the multiple samples

in much the same fashion. We need more info about the methods.

So the exponential rise in odds is one possible scenario. It is based on layers of assumptions founded on unknown, unrevealed information.

[The Doctor.]

79. Therefore you suggest that we can trust no

models at all unless the whole population is represented.

Not only is that impractical, but it is the same argument being used to do away with exit polls entirely.

"The only poll that counts is the one taken by counting all the votes."

[whistle]

13. To micro to be valid...there are 100,000 precincts in the U.S. or

...possibly much more. The selection must be random and of enough precincts to catch at least a standard minimum acceptable example of fraud and should be assigned by state or counties. That would present a much larger sample requirement I would think than now exists.

For example, I live in Orange County Florida where I believe there are about 100 voting precincts (I'm not able to find the exact number). The exit poll samples were done at just eight precincts and they were conducted during an eight hour period even though the polls were open from 7:00AM to 7:00PM. Even the samples were not systematically selected as you suggest in your example, but were set by the polling companies to be a convenience sample of a specified quota, to be obtained in the best and quickest way possible by the field interviewers. That is totally unacceptable as a valid exit sample in my book!

It was done to simply cover people's asses and has no applicability to proving or disproving election fraud or voting violations. These practices are worse than useless because they give people the false impression that all steps were taken to protect the integrity of the voting process, when in fact nothing could be further from the truth.

[BlueEyedSon]

16. BTW, showing one (hand-picked) counter example does not constitute a proof

[Nederland]

18. Agreed

At this point I have merely proved it is possible--a point which, BTW, TIA has disputed. The proof comes next.

[BlueEyedSon]

21. I suspect neither of you have enough data data to "prove" your assertions

Also, TIA has not "proven" anything. My recollection of his posts it that they consisted, primarily, of his computation of the probability of an observed set of events (exit poll results vs actual state totals?).

You could argue that he started with the wrong data/assumptions (garbage in, garbage out) or that he applied the wrong formulae/methods.... or a little bit of both.

If you have some idea where he messed up, why not fix the calculations and tell us YOUR estimate of the probability of the observed events.

:)

[Nederland]

25. Response

I intend to demonstrate that trying verify election results using the availible data is physically impossible. Therefore, it would not make sense for me to "fix the calculations and tell us estimate of the probability of the observed events."

[BlueEyedSon]

28. OK, how about a range?

[Nederland]

30. Nope

I'm sorry, but plugging numbers into equations that rely on complete unknowns is a waste of time. I will explain that more fully in Part II.

[BlueEyedSon]

32. You're not really a statistician, are you?

[Nederland]

37. Here you go

The odds of the exit poll results being accurate are between 0% and 100%.

Happy?

[BlueEyedSon]

41. Hope your "part II" is more elucidating than that.

[Nederland]

47. Ask a stupid question

...get a stupid answer.

Its not my fault you don't understand the concept of a value being "unknown".

[johnaries]

27. You make a valid point, Statistics never "proves" anything.

The only thing that has been proven is that Either the Exit Polls were wrong or the Ballot Count was wrong. What I believe TIA has "proven" is that there is very good reason to suspect the Ballot Count wrong and should be investgated further.

[garybeck]

22. Exit poll debaters: read this article

This article makes some valid points about exit polling.

My personal view is that the exit polls can be seen as a red flag. But they would probably not up in court as proof of fraud because of the points made in this article.

Fitrakis has stated that one reason there is an attempt to punish them for making a frivilous lawsuit in Ohio is that they based their argument on exit polls.

I support TIA's work, and I think it's good for people to look at the data and discuss it. But I think if we are going to win any court battles, we need to pry open those machines. Arguing about exit polls will not win court cases. Please read the article here with the link below.


http://www.ecotalk.org/ExitPollMadness.htm

[Nederland]

24. Good article (nt)

[papau]

36. Garbage in garbage out is the only valid point in article -and the US exit

polls have not been shown to be garbage in.

[John Q. Citizen]

40. The point the article makes is since the exit poll methods are secret,

so, why are exit polls any more reliable than secrectly counting the ballots?

How do we know if the US exit polls are garbage in or not?

We have no way of knowing.

I think the article points out the dangerers of assuming unverified evidence is true.

It may well be true that both the official vote count and the exit polls are bogus.

How would we know?

[papau]

55. Exit poll is not "secret" -it just costs an arm &a leg that DUers can not

pay.

[Nederland]

59. We should raise the money

What does it cost?

[papau]

60. $40,000 per state was a quote a few months ago - but I have not

checked since Dec 2004 - and before the end of Feb 05 the results were not available.

You would think the NY Times would pop for the money and get an interesting story out of it.

And the Strat model was discussed in confidence a few weeks back - but it was only a panel and I do not know if there were take aways.

The fact that I see zero discussion now about the strat model is curious (even from Zogby). It is like the industry has been told to not check for fraud because the industry would be hurt if it did. "in confidence" model should not prevent peer review with a reasonable time period for that review.

Or perhaps my tinfoil hat is on too tight again :-)

[Chi]

48. Question

Heya Nederland (waves)

Ohio has 11,360 voting precincts, would you agree we could find precincts that would match the final state tally down to the tenth?

[Nederland]

51. Not sure

I would say that obviously the best answer would be to have Mitofsky release all of his fucking data so we wouldn't have to guess at these things.

[Chi]

53. Why highlight that in your initial example if you don't know?

"There isn't a single precinct that the polling firm could have chosen to get results that would be accurate for the entire election."


I must be missing your logic here.

[Nederland]

61. That statement refers to the sample data

...not the actual election.

[Chi]

72. I have no idea what you mean, but thats OK....

As long as we both agree that
an absence of precincts, representative of the state vote tally, was not an issue...
Then we're on the same side of the fence anyway.

[L. Coyote]

74. Finding those precincts would be easy.

But what you need to find to do a valid exit poll is a representative sample of precincts, not the ones that matched a particular outcome. A representative sample will include 90% Rep precincts (if there are any) and 95% Dem precincts (these exist), etc. etc.

[Chi]

84. Understood and agreed.

Heya L. Coyote (waves)
My statement was not correct.

Though my point still stands.
Precincts that accurately represented the area mitofsky was trying to poll, did exist.
Absence or availability of them was not an issue.

[anaxarchos]

49. Please cite a single serious source...

You are not attacking TIA. You are attacking (explicitly in other posts) the possibility that exit polls can generate a random sample. That is either a specific criticism or a general one. If it is a specific criticism, please cite a detailed and specific criticism of NEP (not a blogger, not any "mystery" guys, not Mistwell or mgr, not a hypothesis or opinion)which proves that contention. I am not aware of one. The NEP commentary argues the opposite.

If on the other hand, you have a general criticism of exit polls or of Mitofsky methodology specifically, that is a scientific criticism. Please cite a serious academic critique (one that stands up and does not come from the web or "Oral Roberts University") that supports your contention that exit polls can not produce a random sample (not how it might be done differently or "better" or for different design objectives).

This is not a "soft issue" like "will the Mets win the World series". Exit polls are commercially and academically accepted and live (uncomfortably, sometimes) in the world of science.

Until then, this is just another hanging chad.

[Nederland]

50. Wrong

I am not attacking the possibility that exit polls can generate a random sample. They certainly can. I am attacking the idea that with the information currently availible to us today that you can determine whether or not the sample was representative. In Part II I will further assert that if fraud occurs it becomes impossible to determine if the sample was representative.

[kobeisguilty]

56. lets see this part II already

[Nederland]

57. It up now (nt)

[anaxarchos]

69. Nederland, I stipulate that with exit polls in elections with...

... a serious suspicion of fraud, it is difficult, prima facie, to "verify" the randomness of the exit poll. It is hard to tell exactly what is at fault (the election or the exit poll). That is why the postmortem is so important (as it was in this election). It does not follow that nothing can be concluded because the relative weight of some factors are intially unknown. That is the starting point of the analysis and not the end of it.

I read your part II. My request for citation stands.

[L. Coyote]

75. Like Enron, polls in American politics "live" in the world of business

In the world of science, where I "live," methods are transparent and results are peer-reviewed.

Exit polling is just another American business. I will only admit them to my realm if they wish to follow the rules of science and academia. That means total transparency. Until then, I'll accord them the same amount of trust I accord to Enron, Halliburton, and a Bush oil company.

If the pollsters would rather be in the company of academicians, they know what to do. If transparency served their ends (profit), they would do it.

Transparency might put them out of business! What are they hiding and why are they hiding it?

[anaxarchos]

81. I also have "lived" in that world...

L.Coyote, I agree with the gist of your response. It is not a secret that the commercial exit pollsters "short circuit" what formally might be called a "proper methodology". It is also no secret that they do this in order to minimize their costs and make their "product" commercially viable for their "clients". Yet, none of that has led "the company of academicians" to dismiss commercial exit polls as a whole. There have been many citations on these boards which complain about "the black art" aspects or argue with the elements of method, but there is also a begrudging acceptance of the "craft".

To do otherwise would be to deny, on the face of it, all commercial research which doesn't conform precisely to academic research standards and to reject all data which carries a commercial bias (drug studies come to mind). In fact, scientific standards don't set the boundaries for "truth"; they merely state a preference for how to work with evidence in order to minimize the bias which exists both inside and outside the ivory tower.

It is one thing to say that you would prefer that Boeing engineers adopt a "more scientific" methodology, without commercially motivated "short cuts", and entirely another to say that what they build can not fly. Looking up into the sky resolves that debate.

Finally, there is also a problem of "timing". If one has a critique of chess, that the game is really a game of chance, etc., it behooves that person to state those criticisms before one sits down to play. If one plays, boasting all along about their skill and knowledge of the game, and then knocks over the chess board when the negative outcome of the game becomes apparent, a serious credibility problem is created. It is then a little late to present an in depth criticism of the game itself.

So too with exit polls... It has now been a five month long debate and the bulk of the commentary from those who do not believe that the exit polls indicated fraud has focused on the "misunderstanding" of the methodology of NEP and support for Mitofsky's credentials, etc. That shifted dramatically after the Edison/Mitofsky postmortem was released. When that postmortem failed to take flight, the transformation of the criticism to a direct attack on exit polling itself raises the specter of political bias, a bias which in this case is even more obvious than commercial bias.

In a phrase, at this point it just ain't gonna be that easy.

[T Town Jake]

64. Well, among thinking people...

...TIA's "analysis" and horseshit threads have long been seen as something of a joke. But thanks for taking the time/effort/flames-you'll-no-doubt-get to meticulously point it out by the numbers, so to speak.

[anaxarchos]

65. Listen to yourself... that sure helped... n/t

[kobeisguilty]

66. its posts like this

that make me wish DU had a moderator system like slashdot.

[Name removed]

67. Deleted message

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

[kobeisguilty]

68. thank you for making my point :)

"DU signs up dozens of new members every day, and 95% of them are legitimate, concerned liberals. When we make fun of people because of post count or insinuate newbies are disruptors or tell a newbie their opinion doesn't matter because they only have 50 posts, we drive away the future of DU. This behavior must stop immediately."

"If you see someone engaging in this type of personal attack, please hit 'Alert' so the moderators can deal with it. Thank you."

- http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x123856

[T Town Jake]

70. Then do your do duty and hit "alert"...

...don't see why you're still wasting valuable time talking to the likes of moi.

[kobeisguilty]

71. all i meant by the moderator comment

is that you can express yourself without belittling other people's viewpoints, thus posts that are nothing more than personal attacks can be hidden from view.

[L. Coyote]

76. There is a history here at DU and the comments made reflect

a knowledge of that history and have context unknown to a newbie. TIA was even banned at one point. So there is moderator impact. But not every message needs to be read before going online.

This is spinning out in part because the context of the original comment is not known to newbies.

I gave up on critcizing TIA's methods and assumptions a long time ago. Critical comment seemed to lead to thread abandonment and a new thread taking its place. Deaf ears and hide it from other ears too! Now I just skip over ALL CAPS threads w/o seeing who started the thread (that's history and context at work).

This discussion of assumptions and unknowns is nothing new.

[Bill Bored]

78. Damn, after 1000 posts, I finally did an all caps!

Thought it was pretty good too. Even TIA liked it!
Chk it out L. C:

<http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x341483>

[Bill Bored]

77. In fairness to TIA

Mitofsky said his precinct selections were fine. He blamed the discrepancies on within-precinct-error (WPE).

[Freddie Stubbs]

80. Mitofsky didn't blame it on election fraud?

:shrug:

[Nederland]

82. Yes he did

He said that his precinct selections were fine, and then he used the recorded vote to weight the sample vote to produce his final exit poll results. TIA rejects this methodolgy on the basis that you can't use corrupted recorded vote totals to weight the exit poll samples. Unfortunately however, you don't have any other choice. If the recorded vote is fraudulant, it is impossible to verify that the exit poll sample is representative.

[Igel]

83. I think he and TIA are talking in different directions.

Mitofsky was talking about why his final numbers were off, and trying to figure out how that happened. I doubt he was overly concerned with why numbers he didn't authorize the release of were wrong.