I'm currently working with a Dirichlet multinomial model where I wish to integrate out the parameters of the multinomial model. This would be most efficient by having access to a Dirichlet compound multinomial distribution. Is there any chance this will ever become part of JAGS?
Note: I meant to post this in the Help forum
We already have a Beta-Binomial distribution, so I suppose a Dirichlet-Multinomial could be a generalization of that.
Does this imply that the Dirichlet-Multinomial will be part of a future distribution?
It could be if you add a formal feature request (see the menu bar at the top of the page).