Date:         Fri, 11 Feb 2000 09:52:39 -0500
Reply-To:     Julia Hertl <jah12@CORNELL.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Julia Hertl <jah12@CORNELL.EDU>
Subject:      PROC GENMOD question
Content-Type: text/plain; charset="us-ascii"

Hi,
We are analyzing binary data with PROC NLMIXED, PROC MIXED, and PROC
GENMOD. These binary observations (diseases) were taken at several visits
to pigs clustered within herds. We would like to account in our analysis
for correlations among diseases (i.e. spatial clustering) and for
correlations among visits (temporal clustering). Herd will be included in
our model as a random factor.

Our first question is whether it is possible to have a random factor in
PROC GENMOD. We would also like to use observed and predicted values in the
comparison of the results from the 3 procedures (PROC NLMIXED, PROC MIXED,
and PROC GENMOD). We would like to know what the interpretation is for
predicted values in PROC GENMOD, i.e. if herd cannot be specified as random
in GENMOD, is it possible to compare the output from these different
procedures?

Thank you very much.
Julia Hertl
Section of Epidemiology
College of Veterinary Medicine
Cornell University
Ithaca, NY 14853