After the polling fiasco in NH maybe some of you will understand why I post some of the math threads

[dsc]

I post. Olberman had a great segment about the polls with some information which hadn't been posted in the past. It helps explain some of what happened.

First, evidently the polls had Obama's percentages correct, or at least well within the MOE.

Second, there were huge numbers of undecideds and they evidently largely went to Clinton. In most cases, Clinton's percentage plus the undecideds was a bit less than her total numbers.

We often take these polls as vastly more precise than they really are. A poll showing Huckabee at 42 and Clinton at 49 is actually statistically tied, not a Clinton lead. I know that is counter intuitive, but it is the way it is. The 'poll of polls' should be more accurate but we still have a MOE and we still have undecided voters which will eventually make up their minds.

Proving a negative is always impossible. But the math appears to show a break for Hillary at the last minute, just like the math in Iowa showed a break of Obama in the last days.

[ORDagnabbit]

1. so can you explain why

why they seem to only poll correctly in hand counted areas.

polling to hand counts = correct
exit polling to hand counts = correct


polling to machine counts = error
exit polling to machine counts = error

[dsc]

2. actually they didn't poll correct in either place

and also the machine places were cities, where Clinton did better both in the preelection polls (look at your very own numbers) and on election night. But it should be noted that the vote counts are well within any margin of error from the exit polls.

[ORDagnabbit]

4. the numbers flipped only in the machine counted areas......

obamas and clintons numbers switched places when the machines counted them and that doesnt seem odd to you?

[dsc]

5. No because there is a huge confounding variable

She, by every and all accounts, did better in all polls in the places the machines were.

[ORDagnabbit]

9. soooo that being the case why isnt the exit polling and regular polling

putting her at close to 40% like the machine count gave her.

[dsc]

11. the regular polls were wrong

and I provided a clear, concise, easily understood explanation. The undecided broke her way. The exit polls were within the MOE.

[sandnsea]

6. That was an early exit poll

Matthews just said he got the note at 5:30. Usually women vote heavy during the day and men turn out after work. If they thought men would show the way women did, and go for Obama, then I can see where the exit poll was wrong. 57% of the Republican vote was men. Way more women than men voted. The women went for Hillary, the men for McCain. It's really simple.

[ORDagnabbit]

8. no no no no....sorry to rain on the parade 8:10pm is not early

http://blogs.usatoday.com/onpolitics/2008/01/exit-poll-obama.html
Sen. Barack Obama: 39%.

[sandnsea]

10. Released at 8:10 - it's 39-38???

What are you complaining about? That's accurate. It was too close to call. The kids didn't show, women went for Hillary in unpredictable proportions.

[ORDagnabbit]

12. take a look at the chart again.....big variance in machine counted ballots

who supports transparent and verifiable elections.

polling to hand counts = correct
exit polling to hand counts = correct


polling to machine counts = error
exit polling to machine counts = error

[dsc]

14. The margin of error of polls is 3%

that makes all the errors in Obama's vote, every single last error in Obama's vote, well within the MOE.

[ORDagnabbit]

15. but only in the machine counted areas. The hand counted ballots match the polling

[dsc]

18. I honestly don't know how else to explain this

I will try this one more time. Polls have a built in margin of error. This MOE is determined by a fairly simple mathematical formula. Due to the twin factors of expense and speed (polling slowly ads other problems) the sample size of the polls is usually limited to a size which would lead to a MOE of 3%. Thus when a poll says Hillary has the support of 35% of the voters it really means that 95% of the time Hillary's true level of support will be captured by the range of 32% to 38%. Exit polls have fewer non math problems and thus are typically more accurate. There are no undecideds, there are no people claiming to vote when they won't vote, and the poll can be designed to more closely match those who turn out. But even with all of that, there is still a MOE and thus the percent in a poll isn't really a fixed number. Therefore, all of the error, every last bit of the error, is accounted for by the MOE of the exit poll. I really don't know what else to say.

[ORDagnabbit]

19. how bout recount? :)

seriously though every single ballot in new hampshire is paper and we have the perfect opportunity to verify how well the machine tabulations are doing.

I dont buy the MOE and you dont buy my figures but we could both by into the creation of a transparent, verifiable election process by comparing the hand ballots to the machine tabulators.

[dsc]

20. I would have no problem with one

but I don't think you would be any happier with the recound.

[ORDagnabbit]

21. I would be because I'm a kucinich supporter and care about verifiable and transparent elections for

everyone not just my candidate.

http://www.bradblog.com/?p=5532

BLOGGED BY Brad Friedman ON 1/9/2008 4:35PM

Tribune Media: MSM Failed to Note 'Hackable Diebold Red Flags' in New Hampshire Primary Results

Syndicated Columnist Robert Koehler Writes for Tomorrow's Papers: 'Possibility of Tainted Results, a Prospect Most of Media Can't Bear'

Notes Problems With NH's Diebold Machines 'Remain Unsolved'...

Syndicated Tribune Media Services columnist Bob Koehler bumps up our serious concerns about last night's wholly untransparent, and still-uncounted (by anything but a hackable Diebold computer, and a company with a questionable past, to say the least) New Hampshire Primary election results, from "blogger conspiracy theory" to mainstream media concern.

[sandnsea]

17. Dixville Notch is a hand count

Should we base the entire state on Dixville Notch? They went Clark in 2004, the cities didn't. Did Kerry steal NH too?

[ORDagnabbit]

13. heres another one out at 8:10 pm from faux news...blech

Obama 39%
Clinton 34%
Edwards 18%

http://blogs.guardian.co.uk/usa/2008/01/exit_polls_obama_and_mccain_ah.html

[sandnsea]

16. Released at 8:10, because the polls closed

They wouldn't have had time to do the calculations from a poll taken at 8:00. They release exit polls AFTER polls close, that's what those times mean.

[sjdnb]

3. I always go to Political Arithmetik

http://politicalarithmetik.blogspot.com/2008/01/polling-errors-in-new-hampshire.html

While the math behind the analysis may be complicated, the explanations and graphics are not.

[dsc]

7. great site

I will be using this next year in my stats class.