[R] glmm multinomial?

[Prof Brian Ripley]

On Fri, 14 Jan 2005, ecatchpole wrote:

> I'm looking for something like Brian Ripley's glmmPQL that will handle 
> multinomial data. Does anyone know of anything?

It's a lot more complicated conceptually.  A multinomial model has K-1 
linear predictors which should probably have a correlated joint 
distribution.  We don't even have software for multivariate linear mixed 
models, AFAIK.

Would a surrogate Poisson model be good enough? glmmPQL can fit those.

-- 
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