[R] bayesian mixed logit

[Ben Bolker]

Carlo F <c.fezzi <at> uea.ac.uk> writes:

> I am tryin on simulated data, say 300 repondents, 8 choice per person, only
> 2 random parameters... so i think it should be well behaved... so I am
> probably doing a mistake! 
> 
> Thank for poiting out the MCMCglmm package, ultimately I need to write my
> own non-linear untility function, so I do not think I can use standard
> packages hence I am writing my own code... you know if MCMCglmm can deal
> with non-linear in parameters utilities,

  It can't.
  But it still might be useful to cross-check with the MCMCglmm results.
You might be surprised at how big a data set you need to get identifiable
random effects ...

   Ben Bolker