02-06-2016 03:00 PM
I am using SAS 9.4 to run a model using PROC GEE with a multinomial outcome (3-levels), binary and continous independent variables, and repeated measures. I have a significant interaction term between two categorical variables.
Typically I would use the estimate or lsmeans statement to calculate the odds ratios for the possible combinations, however these options are not available with PROC GEE.
How can I generate the standard errors/confidence limits for each of the combinations of the interaction term? Is there another way to code my interaction term?
PROC GEE data=dataset descending; class var1 var2 var3 ID/Ref=first; model outcome= time var1 var2 var3 var2*var3 / dist=multinomial link=glogit; repeated subject=ID /Type=ind; run;