Date:         Thu, 13 May 1999 14:38:07 EDT
Reply-To:     Haiyi Xie <Haiyi.Xie@DARTMOUTH.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Haiyi Xie <Haiyi.Xie@DARTMOUTH.EDU>
Subject:      Re: generalized estimating equations

--- Katherine M Mcknight wrote:
I am wondering if anyone out there can explain the difference
between generalized estimating equations (GEE's) and random effects
models. In my very limited reading regarding GEE's, it appears that they
are very similar to random effects models, and I'm not clear on how they
differ.
--- end of quote ---
 
 
Some of the differences:
 
(1) GEE: population average model; random effect model (REM): subject specific
model.
 
(2) GEE at least GEE1 treats covariance structure as "nusaince"; but in REM or
mixed model, you have to model covariance structure correctly.
 
(3) GEE is based on assumption of missing completely at random, MCAR (there are
new development in this regard recently though);  but REM is based on
assumption of missing at random (MAR).
 
Haiyi Xie