This data set below is part of a larger set for an experiment to compare 2 treatments (GROUP) and the outcome variable is millimeter measurements (DEPTH) made on teeth nested within subjects. The measurements were made at 4 time points (WEEKS), and there are some missing values for DEPTH. Also, although most subjects have 2 treated teeth (one for each treatment), some subjects have more than 2 treated teeth. I would like to fit a repeated measures model to compare the 2 treatment groups, and later I may include other covariates in the model. Is the following model adequate? If not, I would appreciate it if someone would suggest applicable modifications.
Thanks you.
The data set (partial):
Subj Tooth Group Depth Weeks
1 AA A 3 BASELINE
1 AA A 0 6
1 AA A . 12
1 AA A . 24
1 BB B 4 BASELINE
1 BB B 3 6
1 BB B . 12
1 BB B . 24
2 CC B 2 BASELINE
2 CC B 0 6
2 CC B 0 12
2 CC B 0 24
2 AA A 3 BASELINE
2 AA A 0 6
2 AA A 0 12
2 AA A 0 24
2 GG A 5 BASELINE
2 GG A 1 6
2 GG A 1 12
2 GG A 2 24
The proposed model:
PROC MIXED data=long;
Class subj tooth group;
Model depth = group | weeks;
Repeated weeks / subject=tooth(subj) Group=group ;
Run;
