03-13-2018 04:58 AM
How does PROC HPSPLIT try candidate splits when I use the CLASS statement to specify a categorical variable?
I have a variable that has values A, B, C,... etc. The values do not have any ordering to each other (they are nominal). Hence I want a split
to be as common as
But I suspect that there are some bias to to get alphabetical ordered groups in the split.
I stumbled across that i can change from
CLASS myNomVar; MODEL NumTarget = myNomVar;
INPUT myNomVar / LEVEL=NOM; TARGET NumTarget /LEVEL=INT;
Is there supposed to be a different between the two ways of describing the model? But in 14.3 documentation (which I usually use as a reference) those statements are removed? What does that mean?
I think I am on 9.04 TS1M4, but I don't fully understand the versioning in SAS so I might be misstaken.