MMRM in SAS

Mixed Models

Fitting the MMRM in SAS

In SAS the following code was used (assessments at avisitn=0 should also be removed from the response variable):

proc mixed data=adlbh;
  where base ne . and avisitn not in (., 99);
  class usubjid trtpn(ref="0") avisitn;
  by paramcd param;
  model chg=base trtpn avisitn  trtpn*avisitn / solution cl alpha=0.05 ddfm=KR;
  repeated avisitn/subject=usubjid type=&covar;
  lsmeans trtpn * avisitn / diff cl slice=avisitn;
  lsmeans trtpn / diff cl;
run;

where the macro variable covar could be UN, CS or AR(1). The results were stored in .csv files that were post-processed in R and compared with the results from R.