|
"On top of that, I think there was also a question of the confidence interval. Mystery Pollster says you must get to a 99.5% confidence interval and (I am guessing here) I surmise Freeman was only able to get to 95%. How you get from 95% to 99.5% I'm not quite sure, other than to have a much larger sample, which may not be possible to get if enough people were not exit polled"
That is patently untrue. The standard confidence interval used in statistics is 95%. That means that the population mean will be within 2 standard deviations from the mean 95% of the time.
Pollsters NEVER use 99% confidence intervals (3 standard deviations) around the sample mean because the confidence interval around the mean would be too large to be meaningful.
When you have a 95% confidence interval and 5% tails, the probability the actual population mean would deviate from the MOE IN ANY GIVEN STATE is 2.50% for Bush and 2.50% for Kerry.
For Bush the actual vote deviated beyond the MOE 16 times. That means Bush beat 40 to 1 odds in 16 out of the 41 states which deviated in his favor.
For Kerry, this did Not happen even ONCE in the 10 states which deviated in his favor.
To declare that the 99% confidence interval should be used is a canard which has no basis in actual polling. Pollsters ALWAYS use the 95% confidence interval in their work.
Hope that clears it up.
|