William Shannon asked:
What happens if you treat the DV as continuous?
Regression trees did not work particularly well, although they worked better than a classification tree on the whole data set. I haven't tried ordinary least squares regression, but could do so. One problem is that we eventually want to use this to classify people. We could assign people to the class nearest their predicted value, but I'm not sure if that's right. The scale is very widely used, and it is not entirely clear that there is a continuous scale underlying the classes - although it seems likely that there is.
There is also work by Marilyn Ritchie at vanderbilt (I assume genetics) and Rob Culverhouse (cced on this) in my group that uses a partitioning algorithm (not tree but perhaps usable) you could try.
Thanks! I will google their names.