Date:         Thu, 8 Sep 2005 16:01:56 -0400
Reply-To:     SUBSCRIBE SAS-L Anonymous <YTAmherst@MAILCITY.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         SUBSCRIBE SAS-L Anonymous <YTAmherst@MAILCITY.COM>
Subject:      Single-Observation Cluster Problem in Computing GEE Estimates

Dear All:

I came across this message via the following link,
http://www.ssc.wisc.edu/sscc/ssccnews/oct97

--------------------------------------------------
Bug Discovered in Software Computing GEE Estimates
--------------------------------------------------
A bug has been discovered in software which computes GEE estimates.  The
GEE estimates are incorrect when single-observation clusters exist
(singletons) AND you have specified an exchangeable correlation matrix.
This bug affects the GENMOD procedure in SAS, the Oswald command in S-Plus,
and the xtgee command in STATA.  For more information about the problem
and possible solution, see http://biosun1.harvard.edu/~carey/gee.html.
More information is also available from the software vendors on their web
pages.



Does any of you know if this problem has been solved? There are single-
observation clusters in my data and I have specified an exchangeable
correlation matrix using PROC GENMOD.