Thursday, March 27, 2008
Who Are You Going to Believe, Me or Your Lying Eyes...?
Earlier, Howling Latina wrote about the NBC/Wall Street Journal poll. Irrespective of what spin Obama supporters may try to put on the poll, here's a little gem from Taylor Marsh that should put the oversampling issue to rest.
The howler stands by her post. Allowing African Americans (a group supporting Obama by 90 percent) to make up 25 percent of a poll sample to test whether voters became disaffected after the Wright incident is dishonest.
I’m surprised they released this poll. The actual results mention the over sampling but it doesn’t match up with the question about race. This shows 11% were AA which would be 77 out of 700. But, they state they oversampled by 100 which means 177 were AA (25%). I don’t see how they can make any inferences about overall voters since the sample sizes aren’t reflective of the actual population. They don’t say what they did with the extra 100 AA respondents. Plus, the AA respondent results have a higher error (since it’s a smaller sample size) so you really can’t say anything about those results. (This poll's results has an error of +/- 7.4%!]
But she's mighty glad to see the survey mirrors Democratic primary voters by allowing Republicans to vote in the poll. As Marsh writes, this "poll should never have been released."