Monday, October 3, 2011

bad statistics

I was thinking today in my econometrics class that perhaps, a few years down the road, I would like to teach a course in bad econometrics. There would be a unit on data-mining (overfitting), certainly, and multiple comparisons; there could be some coverage of some of the weak IV literature from the '90's; cointegration, heteroskedasticity, colinearity, and other misspecification problems; and certainly the prosaic beginner-level material on comparing data series that have been made incomparable, for example by deflating with different inflation indexes. The hope, of course, is that students would learn to recognize and avoid these errors, but it would appeal to my temperament to try to present it "with a straight face", as it were.

If anyone else has any other suggestions, please offer them.


dWj said...

Growth in the M1 money supply causes the purchase of Christmas trees.

dWj said...