@tarheel did most of the heavy lifting. I just packaged it up with a bow. Here’s the discrete-continuous multi-variate gaussian case. It uses the truncated MVN to derive the bounds from the latent continuous margins needed for the discrete margins.
It’s in PR status but you can check it out at
added mixed_copula by spinkney · Pull Request #13 · spinkney/helpful_stan_functions (github.com)
You should be able to add gaussian, bernoulli, and poisson or leave any one of those out and the code should run. If you want to add different margins it’s a bit of work but do-able. Ideally there’d be a functor that could take in the margin distribution and derive outcome. Until then you’ll have to add it - pr’s are welcome in the helpful functions repo :)