CLUSTER AROUND, FOLKS. CHOOSE YOUR OWN MOE.

[TruthIsAll]

There has been some questioning of my exit poll probability
assumptions. Some DUers and other blogs have claimed that
there is a "cluster" effect which inflates the exit
polling margin of error beyond that of standard pre-election
polls.

I have always believed that exit polls are more accurate than
pre-election polls, for three basic reasons:

1) The exit poll sample size is typically much higher than
that of pre-election polls, by a factor of three or more.

2) People polled after voting know exactly who they voted for,
whereas those polled prior to the election may change their
mind or  not vote at all.

3) Exit pollsters have performed admirably in their work for
the media over the last 30 years or so. The media keeps paying
them, so they must be doing something right. At least until
this election.

Anyway, here's a Christmas gift for those who believe that
Exit polls are, shall we say, suspect.

- Pick your favorite Exit Poll unweighted average MOE and look
up the corresponding number of states (and the odds) of Bush
exceeding the MOE from the exit polls to the vote.

- Pick your favorite cluster effect. You know, the percentage
increase in the calculated MOE due to clustering. It seems
there is a lot of controversy about this percentage. Some
think its 30%, some 50%... and some think its bogus.

- Choose the odds that Bush could exceed his exit poll
numbers. This way you can look up the cluster effect to
justify your choice.

Be careful. Don't assume an (unweighted) average MOE below
2.75. The odds are one out of almost INFINITY. And we all know
that NOTHING is infinite. Not even the universe.

I'll stick with the calculated MOE (ZERO cluster). 
I like those numbers. No fudge.


N= number of states exceeding MOE for Bush		
CE = Cluster effect (zero for the calculated 2.85% MOE)			
MOE = Exit Poll Margin of Error (unweighted 51 state average) 
Odds = 1 out of ?

CE	MOE	N	ODDS
			
100%	5.70	1	1
89%	5.40	2	3
68%	4.80	3	7
49%	4.25	4	26
40%	4.00	6	597
35%	3.85	7	3,715
37%	3.90	8	26,885
35%	3.85	9	223,016
32%	3.75	10	2,098,096
27%	3.62	11	22,192,000
23%	3.50	12	262,019,924
14%	3.25	13	3,432,782,579
5%	3.00	14	49,652,431,051
3%	2.93	15	788,997,832,405

0%     2.85	16	13,544,660,533,445

-4%	2.75	17	200,159,983,438,689
-6%	2.67	18	closer to INFINITY
-12%	2.50	20	closer to INFINITY
-21%	2.25	24	closer to INFINITY
-30%	2.00	26	closer to INFINITY
-39%	1.75	33	closer to INFINITY
-47%	1.50	38	closer to INFINITY
-65%	1.00	43	even closer....

[autorank]

1. Kick -- a fair challenge.

:kick:

[anamandujano]

2. one more thing about exit polls

the interviewers are less able to choose/limit the interviewees and thus get the answers they want

[bush_is_wacko]

3. Numbers give me a headache!

I'll leave the numbers thing to those who need it to believe election fraud did or did not occur. I can't beleive there are many out there at this point that are still questioning rather it did or did not occur. IT DID OCCUR AND I HAVE NO DOUBT. What I'd really like to see is the whistleblowers. There are many of us that can only take the mathmaticians and statisticians word for it. I am one of those people.

[Career Prole]

4. I admire your work very, very much

And I appreciate it even more.
:yourock:

[TruthIsAll]

5. This magic MOEment

in time.

Jan. 6, 2005 - (D)emocracy day

[davidgmills]

6. Shall I post

this on Mystery Pollster or should you?

[TruthIsAll]

8. You are much more of a diplomat that I am.

You have inspired this post.
Therefore, your offer to deliver this to the Mystery Pollster is accepted.

The beauty of this table is that one can be very arbitrary in their choice of MOE to suit whatever agenda they may have.

Its like figuring out your tax deductions.

If 30% isn't enough, give 'em 50% or 90%.
Just enough rope to hang themselves.

[davidgmills]

10. will do

Think I will go to Rick Brady (stones cry out)
as well as he is your true nemisis.

Gosh, to think I am a diplomat. I never knew I had it in me.

[davidgmills]

14. The deed is done

One more thing. You have yet to endorse my proposal for a re-exit poll of Ohio.

I know you think it is not necessary, that the proof exists already.

But most legislators are lawyers. All judges are lawyers. Let me put it this way -- the vast majority of them are scientifically and mathematically challenged. (My diplomatic self won't let me call my peers stup..).

Anything harder than 2 plus 2 gives them reason to throw up their hands and get off the bench or leave the chambers.

And does the media have any more intelligence?

That's why I think it is necessary to spell it out for them in the biggest letters possible and in a way that even the dumbest of our electorate can understand. Clearly and unequivically.

[TruthIsAll]

15. I have stated there should be a revote - hand counted paper ballots.

I would not trust any poll at this point.

The best poll is a fully audited, fool-proof paper ballot vote.

[davidgmills]

17. Me too but

I think the odds of that are nil. I would like to see you put odds on that one happening.

So what else can we settle for?

[SharonRB]

21. I am sooooo with you!

I've been saying this since right after the election. I think the only way at this point that we're going to get anything halfway accurate is to have a revote in Ohio -- on paper only. The bad part is that there's no guarantee the same people will come out and vote the second time. The good part is that some of the people who were turned away at the polls will have a chance to vote. I know it will never happen, but that's what I'd like to see.

I also say that for future elections, we should dispense with the machines altogether and go back to paper -- with a paper trail and hand counting we can get an accurate count.

[TruthIsAll]

20. Wate for the onslaught of new spin and fog creeping like little cat feet..

.

[davidgmills]

19. TIA you are going to love this -- 2% MOE in Ukraine Exit Polling

One done by a Republican Pollster no less.

http://www.truthout.org/docs_04/122704W.shtml

[keepthemhonest]

7. you are right

they are ignorant.The truth will come out and all your hard work will not be in vane.

Here's to Jan 6, 2005.

Truth is All what are you going to do when you don't have to calculate out these statistical analyses anymore?:toast:

[bear425]

9. TIA, just want to say that I am amazed at the work you do!

I am significantly math/statistics challenged, but I can clearly see that your work shows the probability of a legitimate * win is out of the question.

Have you sent your findings to Arnebeck, Conyers, et. al.? Seems like they would be able to use your research. Again, thank you and happy holidays! :)

[dzika]

11. TIA - I've been wondering how we can...

...preserve and amplify all of the important work you are doing.

Are you reading your PMs these days?

[TruthIsAll]

13. I Bookmark them. The threads are saved in the DU archives.

And I maintain a thread of links which I periodically update and post.

You can always sort by author to get recent posts.
Use Advanced Search to search the archives by author.

[madison2000]

12. For a moment I thought MOE was Means Of Execution for Bush

I think someone should have a poll on this!

:nuke:

I vote we send him to Iraq with the same amount of protective equipment as the average National Guardsman and see how long he lasts!

[davidgmills]

16. Down in Texas

they had death by hanging, firing squad, electrocution, gas chamber, and lethal injection. I understand GWB likes them all.

[corbett]

18. Kick For Kerry!

Numbers don't lie, except at Enron!