[R-sig-ME] Multivariate MCMCglmm with binary and gaussian dependent variables

[Julien Piquet]

Dear MCMglmm,

I just saw a post from 2015 in which a user tried to fit a bivariate
MCMCglmm with one dependent variable from Gaussian family and another one
that wwas binary, so perhaps my question has already been treated but I
couldn't find what I was looking for. Right now, I am trying to run a
multivariate MCMCglmm in which some dependent variables are gaussian and
others are binary, which is exactly the same to the previous so far.
However, in my case I have five variables instead of two (4 Gaussian and
one binary). The thing is that I am not quite an expert in Bayesian
statistics so, right now I've been struggling to find a way to code a
prior for such model. Recently I found an inverse-gamma prior
(prior2=list(R=list(V=diag(5),nu=4.002),G=list(G1=list(V=diag(5),nu=4.002)))
could work (the model ran), but I am not sure whether it is correct to use
it or not. Therefore, I would be really grateful if anyone could help to
know which prior/s I should use for this model. I say priors because
perhaps there are several options and thus model robustness to prior
selection can be assessed. Thank you in advance.

Julien C. Piquet
PhD student - GEEI
Instituto de Productos Naturales y Agrobiología (IPNA-CSIC)
38206 San Cristóbal de La Laguna - SPAIN

	[[alternative HTML version deleted]]