Multivariate normal mixture model in brms


#1

Hello,

I was wondering if it is possible to fit multivariate normal mixture model in brms.

Thanks for your help.


#2

You maan a model with a mixture of two or more multivariate normal distributions?

This is not yet possible in the brms framework as you can only mix univariate distributions.

What you can do is having a multivariate model of multiple (univariate) mixtures.


#3

Thanks for the quick response! Will there be a plan for supporting mixture of multivariate distributions?


#4

Not yet. Feel free to open an isuee on github (https://github.com/paul-buerkner/brms/issues). However, I am unsure if this is ever going to come because the brms framework is not designed for that: A multivariate distribution / family is just the (implicit) combination of univariate families instead of being a family in itself.
Thus, it is conceptually difficult to define a mixture of multivariate families.