05-16-2012 04:13 PM
I am interested in modeling a continuous outcome (CD4 count) measured repeated times (0-7 times) per patient over time, with predictors gender and race. The data is from a complex survey and so sampling weights, clusters and strata information need to be taken into account in addition to the correlation within subjects.
Any ideas on the best statistical method to use?
I am contemplating using proc mixed like so:
proc mixed data=dataset;
class race gender id time;
model cd4=time race gender/solution ddfm=kr;
repeated time/type=cs subject=id r rcorr;
I am just unsure if this adequately takes care of the clustering and correlations.