Tuesday, February 12, 2008

Regression Models! (JM)

This warms my nerdy, statistically-inclined little heart... Brendan Nyhan gives us a pretty fascinating analysis of Obama's support centers. His models show that he tends to do well in small, red states, but much better in the heavily Democratic districts of these states. Keep in mind that the statistical significance of these numbers are not terribly high given the small sample size that is the primary season to date. I think, though, if we are to draw a conclusion it is that red state Democrats have bought in to the Obama is more electable narrative and are desperate to win.

I don't think one can make an predictions based on these models, but it remains the case that people are easily able to fit any causal mechanism they want around quantitative data and we should take most polling and analysis with the grain of salt it deserves. Nevertheless, Nyhan's discussion is interesting, careful not to make any broad pronouncements and absolutely worth a look.

No comments: