Another Poll: Huckabee Not Second In Florida
A new CNN poll of Florida gives a radically different result from another poll released earlier today, which showed Mike Huckabee surging into second place. The CNN numbers: Giuliani 38%, Romney 17%, McCain 11%, Thompson 11%, and Huckabee 9%. The margin of error is ±5.5%.
The other poll, from GOP firm Insider Advantage, had Rudy at 26% and Huckabee with 17%. It had a ±3.8% margin of error. Obviously, at least one of these polls is wrong.
Comments (10)
Allsburg wrote on November 27, 2007 4:56 PM:Interesting. One of the polls is wrong, but Huckabee isn't the telling point. If we imagine that Huckabee actually has the support of 14% of the voters, then that percentage is within both polls' margins of error. It's Rudy's numbers that can't be reconciled: any percentage within the CNN margin of error is outside of the Insider Advantage margin, and vice versa. The margin of error must be larger for at least one of these polls than is reported.
Redshift wrote on November 27, 2007 5:28 PM:Personally, I find the idea of a polling firm associated with the corrupt GOP with the name "Insider Advantage" to be unintentionally hilarious.
cleter wrote on November 27, 2007 6:03 PM:I live in Florida, and the only candidates I see people stumping for are Obama and Paul. A strange mix of libertarians, hippies, and right-wing Christians are coalescing around Paul in rural north Florida.
Randy wrote on November 27, 2007 6:14 PM:I have a problem with each poll. The IA poll took a week to conduct, so it may not really be a "snapshot" in time. The CNN was only in the field for 2 nights, and may not be a big enough picture.
Anonymous wrote on November 27, 2007 6:53 PM:Anybody know how to emigrate to Canada? Giuliani is going to win the nomination one way or another because he has the neocon corruption machine behind him. The Dumbocrats have already allowed the MSM to select Hellary as their candidate. So we're basically screwed.
foo wrote on November 27, 2007 11:39 PM:17% - 9% = 8%
5.5% + 3.8% = 8.8%
the error bars overlap at the 1sigma level. there is no reason to conclude either is wrong. they are mutually consistent, just not sufficiently accurate. ~60% chance of this result given perfect measurement (ie, no biases) with the reported error ranges.
I was going to say what foo said. The poll numbers aren't mutually inconsistent.
Although, it is correct to say that Huckabee is either in second place in Florida, or he isn't.
heretic wrote on November 28, 2007 7:38 AM:Umm. HRC currently does far better than Obama in most state GE poll matchups against Giuliani. It is the nomination of Obama that will leave us screwed.
SMD wrote on November 28, 2007 8:51 AM:Foo and Jim D are incorrect on the statistics. First, there isn't really enough information to answer the question rigorously. But assuming independence and that the margin of error is 2sigma, the standard deviation of the difference between the Huckabee scores in the two polls is 3.3%, and so the difference between the Huckabee scores in the two polls is 8%, plus or minus 6.6%, with 95% confidence. So the true value of the poll discrepancy is as likely to be 1.4% as 14.6%, but it is almost certainly not zero.
foo wrote on November 28, 2007 8:47 PM:re: SMD's comment
It depends, of course, if these are 1 sigma (68% confidence) or 2 sigma (95% confidence) error bars. This is never mentioned in polling articles, but if I look at the sample sizes, the error bars seem consistent with 1 sigma Poisson errors for the sample size (eg, the CNN poll above quotes 5.5% error and 300 people. 1/sqrt(300) = 5.8; I'm assuming the '300' is rounded off from 330). If they are 1 sigma error bars, then the error on the difference is sqrt(5.5^2 + 3.8^2) = 6.6, which means the 95% confidence range is -5.2 to 21.2. If they are 2 sigma error bars, then SMD has the 95% confidence interval correct.
