Wednesday, May 15, 2013

A metaphor for why Goodness of Fit tests are, well, not very good.

I am proud to say I learned my econometric from Art Goldberger, who had little use for R-squared.

Anyway, a smart friend of mine (who works in industry and therefore might not want to be named) pointed out that he could probably fit the brushstrokes of a Jackson Pollack painting with a 17 degree polynomial and get an excellent R-squared.  But he still couldn't predict what a next brush stroke might look like.

No comments: