December 24, 2004

Marriage Gap

Don't get greedy! In my latest VDARE column, I bragged that found that my Average Years Married among white women measure correlated with Bush's share of the vote not just at the extraordinarily high r = 0.91 level (r-squared = 83%) I reported for a linear equation, but at a bizarrely high r = 0.95 (r-squared = 90%) when you take the log of the statistics. A reader writes:

Beware log-log plots; they almost always make correlations look better than they should. You can correlate almost anything log-log. Try taking a shotgun pattern in a target and plotting it log-log. You can sometimes wind up with a nice linear plot.I'd be more comfortable sticking with your puny (!) .91.

I have to admit that I can't think of any reason the log correlation would be more realistic than the linear one. Any thoughts?

My published articles are archived at -- Steve Sailer

No comments: