[R-sig-ME] GEE different working matrices for males and females

[ben pelzer]

Dear list,



I'm estimating logistic models with geepack in R, using geeglm. Is there a
way to specify separate AR1 working matrices for two groups, males and
females? I found an example on how this can be done for "exchangeable"
matrices which differ for males and females. But I cannot figure out how to
do this for different AR1 matrices.



In geepack one can specify a "userdefined" working matrix. The genZcor( )
function can be used to generate the design-matrix for unstructured
correlations, which can then serve as a basis to obtain the design matrix Z
of e.g. an exchangeable structure. This design-matrix can then also be
extended to obtain different exchangeable correlations for males and
females.  The geepack manual also gives an example of how a Toeplitz
design-matrix can be generated, which can then also (although I did not try
that yet) be extended into a separate Toeplitz for males and females. But
there is no example of how to obtain an AR1 design matrix for the
correlations.



Thanks for any help, Ben.

	[[alternative HTML version deleted]]