I am working on a sas macro to perform an all possible regressions
variable selection process for a binary logit model with a single
random effect (using the GEE methodology and GENMOD).
I am currently using PROC SUMMARY with the CHARTYPE option in order to
generate all possible combinations of the potential predictors.
HOWEVER, this only provides all possible main effects models. Any ideas on how I can obtain all possible models INCLUDING higher-order interaction terms (Really: all possible hierarchical models: I want to automatically weed out models that possess higher order interactions without the lower order constituent terms)?