03-03-2017 09:51 AM
Recent trial evaluated each indivdual treatment vs. a control. Seems that adjust=simulate is for all possible comparisons, Dunnett seems to be needed for control vs. each trt. However, is it inappropriate to use simulate if not all-pairwise comparisons? Reason I'm asking is to see if I need to re-run the analysis using Dunnett. Some of the analysis is on blood serum to detect various enzymes that define toxicosis in lambs. Thus, don't want to report "no differences" if there really were but the simulation adjustment was too agressive. Hopefully that makes sense.
CLASS TRT DAY ID;
MODEL serum = TRT|DAY / DDFM = KR SOLUTION;
REPEATED DAY/SUBJECT=ID TYPE = UN;
LSMEANS TRT|DAY / DIFF ADJUST=SIMULATE (REPORT SEED=121211) cl adjdfe=row SLICE=(DAY TRT);
03-03-2017 08:55 PM
I think this pdf "Computation of pvalues for Multiple Comparisons with a Control in the SAS System" will answer your question.