04-28-2016 07:58 AM
In a placebo controlled randomized study invesigating the effect of a drug on weight, blood pressure ect, I found that there is difference in the baseline weight between the two groups (placebo group and active group).
The study has repeated measurements ,and therefore I am using mixed model. To adjust for baseline difference in my outcome weight, I have build this model:
proc mixed data=sasuser.data;
WHERE week > 0;
class id week treat (ref='2');
model weight = week treat treat*week weight0*treat
/solution CL DDFM=KR;
repeated week / subject=id type=un r rcorr;
where weight0 is mean weight at baseline for both groups. As far as I understand, I put it as a covariate (ANCOVA).
However, I want to adjust for other baseline differences, such as age and gender. My question is therefore: How do I do that in the model above? Can I have more than one covariate at the same time?
05-05-2016 09:11 AM
Yes, you may have more than one covariate (although gender would most likely be included in the CLASS statement as a categorical covariate). Get a copy of SAS for Mixed Models, 2nd ed. by Littell et al., and look at Chapter 7 Analysis of Covariance for very good coverage of this subject.
05-06-2016 03:57 PM
You want to use the values of the covariates for each subject (ID), not the group means.
With multiple covariates, I find it useful to remember that my model is now a multiple regression, with all the assumptions and issues associated with multiple regression (multicollinearity, influence, etc.). So in addition to Steve's suggestions (the "bible" for mixed models in SAS, always a good idea), read up on multiple regression.
Walt Stroup told me this week that the 3rd edition could be out this year. Which would be swell.