Yeah - especially when other polls say the opposite - like this:
http://www.usnews.com/blogs/barone/2008/ 03/25/polls-show-obama-damaged-by-revere nd-wright.html
Add that to the fact that MSNBC ha to explain that they OVERSAMPLED AA's in the poll - you know, the people least likely to be affected by the Wright comments and probably the most likely to like Obama's speech and support him anyways?
Once again, oversampling is done to provide statistically significant results on certain sub-groups. The overall results are weighted by the general population, so it has no impact on the top-lines of the poll.
Over samples but not over represented. There is a difference. This is called stratified sampling. It's a good thing to know when you're looking at polls.
Oversampling does not mean what you think it means.
Get your facts straight. They oversample smaller populations so the accuracy of the statistical sample equals the larger groups. They then ADJUST before including them in the full poll results so they are not over represented. NBC confirmed that is in fact what they did.
http://firstread.msnbc.msn.com/archive/2 008/03/27/827746.aspx
not sure how combining the opinions of 100 more AA's into the 77 and adding that result, esp. considering it's got an MOE of over 7%, gives any better result than adding the 177. And how is the higher MOE factored in?
When pollsters poll a population, they don't poll exactly 13% AAs, 11% caucasian Hispanics, 53% women, etc., etc. They poll who they get to a significant sample size, then weight the responses so that the answers they DID get are proportionally represented in the final number.
In this case, they wanted to know what AAs thought in particular, and they needed some minimum number of them to get a number with statistical significance. So they surveyed more of them ON PURPOSE and weighted them less. The adjustment ss the same thing that happens in every poll randomly, it was just forced into the effect.
and the higher MOE?
Not sure what you're asking here -- yes, the smaller sample size has a bigger MoE than the group as a whole, but lower than it would if it were a proportional percentage... which is the whole point of oversampling: to reduce the MoE on your subset.
If your going to troll rate my comment in a different diary, you could at least give an explanation. Nothing I said was inflammatory. Defend yourself if you will.
