Idiots and Statistics

One of the unhappiest marriages of modern technology is that of economists and multiple regression. With the aid of a computer, multiple regression is simple enough for anyone to do it, and far too many of them do. Another egregious example seems to be Dr Dave Berri (with details here).

His technique produces a number of counterintuitive results - for example, he rates Derek Fisher of LA as the least productive player on a winning team. This is sufficient evidence for me to rate Berri as an idiot among economist multi-regressors.

The are several kinds of pitfall for multiple regression, prominent among them neglect of correlations and of variables not in the regression. This sort of thing is especially evident in the kinds of statistics Berri uses. Key stats are what he calls "defensive efficiency" which is essentially just points scoring of the guy you are guarding and "offensive efficiency," scoring per unit time. These stats are very easily contaminated. If you are guarding Steve Nash, Deron Williams and Ray Allen, you will get a lot more points scored against you than if you are guarding some no name, but that just proves that your coach is very impressed with your defensive ability. Kobe Bryant takes a lot of hard shots, but that's mostly because when there are only hard shots to be had, he's the man to take them.

In short, Dr. Berri, there are more things under the Sun than are dreamt of in your lame-o regressions.

Comments

Popular posts from this blog

Anti-Libertarian: re-post

Uneasy Lies The Head

Book Review: Anaximander By Carlo Rovelli