Genralized Esitmating Equations in SAS

#1
I need to determine the effect of one categorical variable on a binary dependent variable while controlling for the effect of a continuous variable. Also my data was obtained on a repeated measures design. I tried to use a binary logistic regression, but was told that this was not correct because of my repeated measures design. My statistics professor suggested I use a Generalized Estimating Equation, but said that was not an analysis that she was familiar with.
As SPSS does not support this analysis I guess I will need to use SAS. However, I have very limited experience with this program and all the explanations of how to conduct A GEE in SAS assume that I know much more about SAS then I do. If anyone could offer or refer me to a sort of dummed down explanation of how to conduct a GEE in SAS that would be great.
Thanks
 

mp83

TS Contributor
#2
I think Agrsesti's book use SAS. So with a bit of a luck you could find it in his site (or even ther's an Appendix in his book,I don't have it at hand this moment)
 
#3
You would use the REPEATED statement in PROC GENMOD and specify a SUBJECT= variable. This variable indicates which subject is which, that is, each observation with the same value of the subject variable is a repeated observation on that subject. You would need a model statement of the form MODEL dependent=independent1 indepedent2... or if you are dealing with a categorical dependent variable, MODEL number of events/number of trials=independent1 indepedent2... When dealing with GEEs, maximum likelihood-based quantities, such as assessment of fit, aren't applicable because the GEE doesn't maximize the likelihood of the model.
 
#4
Your committee member is right, you need to use GEE.

Toxikhan is exactly right about how to do it in SAS. It would be an easy thing to do for someone who knows SAS and understands GEE, but could be frustrating to learn both simultaneously.

However, according to SPSS, GEE is available in SPSS as of version 15.0 in the "advanced statistics" module.

See http://www.spss.com/media/collateral/S17CMP-06081.pdf

If you're generally familiar with SPSS, it would be your best bet, although I have no idea where to find a resource about it.

Karen