On 5/21/08, Sigurd Hermansen <HERMANS1@westat.com> wrote: > No. While aggregation makes data more compact, you don't have to > aggregate into events/trials form. > > SAS documentation shows the response = effects alternative to the > events/trials syntax of the MODEL statement: > MODEL response = < effects > < /options > ; > > as opposed to > > MODEL events/trials = < effects > < /options > ;

Thanks Sig. It seems that when I use the stratification A*B*C to aggregate (sum), and then leave C out of the model, the DF change, given that C has more than one level. I will have to look into the documentation further and hopefully find a book in our library that specifically address poisson regression using GENMOD.

A B C 0 0 0 0 0 1

1 0 0 1 0 1