# Discretizing a continuous variable in SAS using a decision tree

I have a variable X15 that varies from .001 to 10,000 or so and I'd like to construct classes that are unbalanced to another variable, Y, that is either 1 or 0, so that each class will have more of (Y=0) or of (Y=1). All I want is the output from some PROC in 9.3 that will tell me what the bins should be. I've tried PROC SPLIT, PROC DMSPLIT, and PROC ARBORETUM, but PROC PRINTing all the output data sets doesn't give me a clue.

How do I get the splitting values for, say, 5 splits into 6 bins of my X15 data using a decision tree? An arbitrary number of splitting values is preferred, I just can't figure out how to extract that data from a PROC statement.

• Why not use a continuous approach that has much better statistical operating characteristics? – Frank Harrell Jul 29 '13 at 20:35

SAS STAT does not have a PROC that does decision trees. You could export your data to R, which does, or else you need to buy Enterprise Miner in SAS; if you have that, then see e.g. this paper