« Leon Kass on ice cream cones | Main | Distorting McCain on "100 years" in Iraq »

April 08, 2008


I'm very skeptical of these models which predict past elections perfectly yet are constantly updated after each election. Basic signal analysis allows you to find very good fits to this type of data using obviously unrelated variables.

And has anyone actually run the predictions for the last 7 elections based on the original 1978 model?

The comments to this entry are closed.