[R-sig-ME] Fwd: Multinomial response variable

[Ben Bolker]

  I think I replied privately that MCMCglmm can handle multinomial
models, and that library(sos); findFn("{multinomial mixed model}") finds
the mme package ...

  There are no immediate plans to make lme4 do multinomial models. It
can probably be done reasonably straightforwardly by decomposing the
problem into a series of parallel binomial models, but it's enough of a
nuisance and an effort to do it that it's pretty low on the development
list ...

On 16-07-22 05:18 PM, Célia Sofia Moreira wrote:
> Dear All,
> 
> Thank you very much for the lme4 R package. It is really, really useful!
> 
> I have a problem concerning this package. I want to study the following
> model given by the formula:
> 'y ~ time + (1| id)',
> where 'y', 'time', and 'id' are factors, nominal but non ordered, with 4, 2
> and 60 levels, respectively.
> 
> Thus, to perform the model using lme4, I would need 'family=multinomial'.
> However, for the best of my knowledge, this family is not available in
> lme4. Can you please suggest me an alternative option to conduct this
> study? Do you plan to extend lme4 to this family in a near future?
> 
> Kind regards,
>   celia
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>