Tuesday, November 30, 2010

i'm an idiot

OK, this week's poll study results are up. A day late due to Thanksgiving holiday weekend. The usual etc. etc.

There is one major, important change to the numbers over the whole poll. It's like this. The first week I did this, one of the voters' ballots wasn't listed - either he forgot to send it in, or there was a glitch, or whatever. So I averaged by 59 instead of 60. The problem is, I've been doing that the whole time. When I wrote the formula to weight the regional averages, I remembered to add the extra ballot to the SEC total but not the overall total, so all the averages in the "total" column were too high. Overall gist: bias looked stronger than it really was.

It was a quick fix, but it changed the numbers around a bit and on top of that it's bloody embarrassing. Won't affect things much going forward, though, except that if you were hoping for evidence of regional bias, it's a lot less likely to occur.

No comments: