A quick note. I have been able to reproduce Matt's results on his dataset, so I am now investigating what's going on. -- Stephen Skory s at skory.us http://stephenskory.com/ 510.621.3687 (google voice)