Specifying that effects are ordinal?

[JonB_]

Hi all,

I've been asked to build a predictive model with data that includes both continuous and class variables. Some of the class variables are ordinal in nature (essentially self-reported rankings of various things). I'm playing around with both Proc Genmod and Proc Adaptivereg as ways to build my model. How do I specify which ones of my class variables are ordinal? Do I have to code them in a certain way for this to be true?

[plf515]

Have you considered PROC TRANSREG? You can then use the MONOTONE transformation