Date:         Sun, 16 Mar 2008 21:36:24 -0700
Reply-To:     weihai_zhan@YAHOO.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         weihai_zhan@YAHOO.COM
Organization: http://groups.google.com
Subject:      Are there any tricks through which PROC GENMOD of SAS can be used
              to deal with nested correlation for GEEs approach?
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

Hi,

Can anyone help me on this problem?

Setting: Outcome: Hearing loss (0: no hearing loss; 1: hearing loss).
The study includes about 2000 families (every family has one
FAMILY_ID ). Every family includes 1 mom and/or 1 dad, 0-6 offspring.
The mom/dad was repeatedly measured 3 times (in 1994, 1999, and 2004),
while the offspring were only measured once in 2007. Every participant
has one uniform SUBJECT_ID.

Question: How should I deal with the nested correlation issue?

Originally I had thought to use:
PROC GENMOD;
CLASS FAMILY_ID SUBJECT_ID;
MODEL <outcome variable> = <predictors> / DIST = BIN;
REPEATED SUBJECT = SUBJECT_ID (FAMILY_ID) / TYPE = UNSTR;

But I was told that this was incorrect.

Thanks a lot.

Will