[R] glmm with negative binomial

[Prof Brian Ripley]

On Wed, 16 Feb 2005, Brian Aukema wrote:

> At present, can generalized linear mixed models with negative binomial 
> distribution and estimating the shape parameter be fit using R?  I am aware 
> of glm.nb but am wondering about incorporation of mixed effects.

I am not aware of anyone who knows how to do that reasonably robustly. 
If the model were simple enough you might be able to do numerical 
integration well enough to get a likelihood to maximize, but we usually 
have enough trouble with a Poisson model (and our experiments seem to show 
that other people's software has much more trouble).

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595