Date:         Sat, 7 Mar 1998 06:20:44 GMT
Reply-To:     Ming Ji <ez063890@CATBERT.UCDAVIS.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Ming Ji <ez063890@CATBERT.UCDAVIS.EDU>
Organization: University of California, Davis
Subject:      PROC GENMOD question

In GEE, there are a beta-equation for modeling the mean response and a
alpha-equation for modeling the correlation structure. (For example,
in Diggle, Zeger and Liang's 1994 book: Analysis of Longitudinal Data,
pp144-145)  GEE will give consistent estimate for beta even if the
correlation structure (parametrized by both beta and alpha) is not
correctly specified.  PROC GENMOD has an option corrw which can give an
estimate for the working correlation matrix. However if we want to know
how accurate this estimate of the correlation structure is, are there
options in PROC GENMOD to obtain information about the accuracy of
the estimated correlation matrix ?
 
Thanks
 
 --
sincerely yours
 
Ming Ji
Division of Statistics
University of California
Davis, CA 95616
email: mji@wald.ucdavis.edu