03-17-2017 02:03 AM
Is there a way to include 2 class variables in Lasso regression by default? Basically I want to allow Lasso to choose from among a list of continuous variables but want to include 2 class variables (one has 9 categories and the other has 6 categories) for sure. I am using SAS version 9.3 X64_8PRO.
03-17-2017 06:12 AM
Unfortunately, INCLUDE= option is not available for PROC GLMSELECT. But INCLUDE= option is available for PROC HPGENSELECT .
03-17-2017 08:22 AM
You can run a regression on the two variables, then use the residuals as the response in PROC GLMSELECT. It might look something like this:
proc glm data=Have; class C1 C2; model Y = C1 C2; output out=Residuals r=NewY; run; proc glmselect data=Residuals; model NewY = x1 - x1000; ... run;
03-17-2017 03:09 PM