From my model I can get the overall effect of the interaction, but I’m only interested in certain parts of the interaction. I'm wondering if this can be done with ESTIMATE statements.
My data: Each subject was followed for 4 days, and they were randomized to one of two trainers on each day. Subjects were with each trainer at least once, but could have been with one trainer 3 times and the other trainer once (like subject 3). At the end of each day, a test was given.
This is a simplified snapshot of my actual data.
PTNO DAY TRAINER SCORE
1 1 1 50
1 2 2 60
1 3 1 55
1 4 2 45
2 1 2 70
2 2 1 72
2 3 1 90
2 4 2 85
3 1 1 67
3 2 1 79
3 3 1 80
3 4 2 85
4 1 2 62
4 2 1 68
4 3 1 70
4 4 2 72
...
...
From this model I can get the p-value for the interaction of trainer*day, and from LSEMANS / DIFF I can see the comparisons for each trainer and day individually, but I want to know the overall effect of comparing trainers on the same day. Can this be done with estimate statements?
PROC GLIMMIX DATA=ads;
CLASS ptno trainer day;
MODEL score = trainer day trainer*day ;
random int / subject=ptno;
RANDOM day / subject=ptno*trainer type=un residual;
LSMEANS trainer day / cl diff;
LSMEANS trainer*day/ cl slice=day Diff;
Run;
I know how to do an estimate statement to compare trainers on a given day (but this gives same result as LSMEANS anyways so I didn’t include in my model) but I don’t know how to write an estimate statement to compare trainers overall, on the same day (i.e. trainer 1 vs 2 on day 1, and trainer 1 vs 2 on day 2, and trainer 1 vs 2 on day 3, and trainer 1 vs 2 on day 4) Can this be done with estimate statements?
estimate 'Trainer 1 vs 2 on day 2' trainer 1 -1 trainer*day 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
... View more
