Contrast/Estimate Statement to detect specific treatment effect

[suebayran]

I have an GLM and a treament variable given to the patient in a factorial fashion coded in the following manner

1-control

2-oral drug A

3-inhaler B

4-drug A and B togther

My outcome variable is the weekly number of puffs of albuterol taken for the patient's asthma.

I need to test whether the treatment effect, which is significant in my model, can be attributed entirely to the inhaler B. Can someone help me in how I might test this in SAS? Would it be some sort of contrast statement? Thanks all.

[SteveDenham]

Unfortunately, this experimental design isn't really a factorial, as each patient proceeds through the four treatments in the same order.

In that sense, this is a repeated measures design, with a response variable that is a count.  In order to get the comparisons you want, you'll likely want to use PROC GLIMMIX.

Try the following for a comparison of the marginal estimates:

proc glimmix data=yourdata;
class ID treatment;
model count = treatment / dist=poisson;
random treatment / residual type=un subject=ID;
lsmeans treatment / diff;
run;

This should at least get you started.  There may be some glitches along the way as far as convergence, etc., but those can be dealt with.

Steve Denham