The Polls and the Mid Terms

[JohnStOnge]

I keep hearing people say that the polls were all messed up for the mid term elections. So far I don't think so. I've been looking around some and have yet to be able to find a race where the person who won winning was not indicated as reasonably likely to win.

I think people forget about how a sample survey and resulting estimate work. When you see the margin of error, that's a 95% confidence interval. The idea is that before you collect the data, knowing you'll go through the process of constructing a 95% confidence interval, there's a 95% chance that the true percentage is within that interval. I'm not aware of any race in which one would have looked at the 95% confidence interval and said, "This person can't win" then they won.

And even if we could find some races like that, there's another thing to remember: In each individual case where a 95% confidence interval is generated, there is a 5% chance that the true proportion will be outside of that interval. That means that, over a large number of poll results, you'd expect the truth to be outside of the 95% interval about once in each 20 times.

Anyway, the bottom line is that what happened in the mid term elections is not, as far as I can tell, outside of the range of what would reasonably be expected as a possibility based on the pre election polls.

[houndawg]

I keep hearing people say that the polls were all messed up for the mid term elections. So far I don't think so. I've been looking around some and have yet to be able to find a race where the person who won winning was not indicated as reasonably likely to win.

I think people forget about how a sample survey and resulting estimate work. When you see the margin of error, that's a 95% confidence interval. The idea is that before you collect the data, knowing you'll go through the process of constructing a 95% confidence interval, there's a 95% chance that the true percentage is within that interval. I'm not aware of any race in which one would have looked at the 95% confidence interval and said, "This person can't win" then they won.

And even if we could find some races like that, there's another thing to remember: In each individual case where a 95% confidence interval is generated, there is a 5% chance that the true proportion will be outside of that interval. That means that, over a large number of poll results, you'd expect the truth to be outside of the 95% interval about once in each 20 times.

Anyway, the bottom line is that what happened in the mid term elections is not, as far as I can tell, outside of the range of what would reasonably be expected as a possibility based on the pre election polls.

Really? I don't hear anybody saying that.

[andy7171]

NO ONE was predicting Larry Hogan to be the new Maryland Governor.

[93henfan]

NO ONE was predicting Larry Hogan to be the new Maryland Governor.

Andy wins the thread. Nice work, Shrek!

[JohnStOnge]

NO ONE was predicting Larry Hogan to be the new Maryland Governor.

This goes back to understanding how survey statistics work. By convention, the 95 percent confidence level is the point at which you say, "OK. I'm making the decision that it's going to happen."

Check out this site:

http://elections.huffingtonpost.com/pol ... n-vs-brown" onclick="window.open(this.href);return false;

Note what it says over to the right:

CONFIDENCE OF WIN The probability that Brown would beat Hogan was 92%.

In other words, if you were following convention, you would not have made the call that either candidate was going to win. It was in the "uncertain" category.

Besides, as I said, when you're using the 95 percent confidence level you are accepting the idea of being wrong in 1 out of each 20 cases. So a single instance of being wrong doesn't mean the process didn't work.

[JohnStOnge]

Really? I don't hear anybody saying that.

Then you must not have been watching must post election analysis on TV. I've been watching both Fox News and MSNBC in order to get the perspective of the opposite poles and have been seeing /hearing a lot of it.

[houndawg]

Really? I don't hear anybody saying that.

Then you must not have been watching must post election analysis on TV. I've been watching both Fox News and MSNBC in order to get the perspective of the opposite poles and have been seeing /hearing a lot of it.

You should get out more.

[BlueHen86]

Really? I don't hear anybody saying that.

Then you must not have been watching must post election analysis on TV. I've been watching both Fox News and MSNBC in order to get the perspective of the opposite poles and have been seeing /hearing a lot of it.

You are picking the two worst sources of news to make your argument. I didn't watch either one, and most of what I did watch correctly predicted most outcomes last week.

FoxNews and MSNBC are closer to propaganda than they are to news. They don't even try to be fair or balanced.

[Skjellyfetti]