negative binomial in GEE

Tue, Sep 27, 2005 8:23 AM #

Dear R-help,


I was recently wanting to use GEE with the negative binomial "family". It
seems that this is lacking in the otherwise excellent implementations of
the GEE methodology ( packages: gee, yags, geepack).

I would have thought it a simple step to allow the creation of a family,
i.e providing the link function (log mu) and the variance function (mu +
mu^2/theta) , assuming theta is specified; and then to let the gee code
do the business in a similar fashion to glm.

However the gee codes all seem to rely a set of C routines with either the
normal, poisson, or binomial, hardwired in.

Does anyone have any suggestions for a way round this problem for the
future (I had to resort to using Stata), or maybe more realistically, how
much work it would take to build an extendible version of the gee
"algorithm"?


thanks Simon Bond.

Thomas Lumley

Tue, Sep 27, 2005 9:33 AM #

On Tue, 27 Sep 2005, Simon.Bond wrote:

I don't think it would take that much work [and I've done it a couple of 
times, once in XLISP-Stat and once in SPIDA].  An entirely interpreted 
implementation will be relatively slow, though it might be possible to 
improve that quite a bit with sparse matrix techniques.

 	-thomas