I performed a mmrm model as below:
proc mixed data = have;
class id treatment week strata;
model chg = baseline treatment week treatment*week treatment*strata/ ddfm = KR;
repeated week/ subject = id type = UN;
lsmeans treatment*week/ cl alpha = 0.05 diff
ods output Tests3=tests3 lsmeans=lsmeans diffs=diffs;
run;
In the model:
(1) id: represents the patient id, ie 101,102,103....
(2) chg: represents the change from baseline of hba1c
(3) baseline: the value of hba1c at baseline
(4) treatment: contains 2 categories: "drug" and "placebo"
(5) week: contrains 3 levels: week8, week16, week24
(6) strata: represents the stratification by the median of hba1c at baseline, ie:
group1=patient of those who have hba1c at baseline <= median of hba1c at baseline;
group2=patient of those who have hba1c at baseline > median of hba1c at baseline.
In general, I'd like to test the interaction between strata and treatment, but I am not sure if I performed the test in the right way. Because I checked the results (as below), the column of "Diffs in LS means" were pretty close between strata, however the interaction p-value is statistically significant. (The values of Diffs in LS means were got from the output file diffs, and the p for interaction were got from the output file test3.)
I also checked the Diffs in LS means not only at week24, but also week 8 and week16, they all looked pretty close between strata. So did I interpret the data in a wrong way? Or did I perform the model in a wrong way? What data should look different if the p for interaction is significant? Thanks!