Hello. I'm a statistician. I hope you bookmark my page for credible, up to the minute polls and graphs regarding the 2012 race. I'm frustrated with media bias so I thought I'd make this blog for others like me who are as well.
I used to like RealClearPolitics until I realized that they were including polls in their averages that were Democrat or Republican affiliated or included "lead-in questions" which both lead to biased results. I hope my blog will gain enough of a following to influence RCP to follow my lead, but maybe that will never happen.
I am therefore only including polls that fit these criteria:
1. No Democrat/Republican affiliations. This will help to reduce bias.
2. No lead-in questions. If you're only asked how you feel about Candidate X vs Obama and then Candidate Y vs Obama (Candidate Z ignored), and then asked to make your choice, the results will be skewed towards X and Y for psychological reasons, and Candidate Z will be slighted as a result.
3. Publicly available free crosstabs must be available. Sorry, Rasmussen, if you're not going to share your full results with everyone, then that reduces your credibility in my estimation.
4. Only data with a mean polling date within 20 days of the current date will be counted. (Example, if its 11/30, then a 11/7 - 11/13 poll will be counted because its average date is 11/10 which puts it within 20 days. If it was 11/6 - 11/13, its average date would be 11/9.5 which puts it outside of 20 days. That poll wouldn't be counted.)
5. Only phone polls will be counted. Online polls are proven to be less accurate.
6. The averages will be semi-weighted. By that I mean half of the average will be weighted by number of people polled, and half of the average will weight all the polls equally.
7. If both Republicans and Republican-leaning Independents are included in a poll, I will include both because the average will be more representative of who will be showing up at primaries.
No comments:
Post a Comment