Multinomial logistic regression with random effects

[ekbrown77]

What's the best package/way to run a multinomial logistic regression with two random effects? I have a categorical dependent variable with three levels, a handful of fixed effects (both continuous and categorical), and two random effects (speaker and lexical item).

lme4::glmer()

has a

family

argument, but it doesn't take "multinominal" (despite taking "binomial").

I'm wading into waters I haven't tread before.

Thanks.

[Stefan Th. Gries]

My guess would be MCMCglmm but that package is not for the faint-hearted ... (and that includes me).