Conditional independence while sampling using NUTS

General
1

I am trying to sample from a posterior distribution. The model is defined by a DAG of parameters. There are several conditional independencies among the parameters to be sampled. Does Stan automatically take advantage of this to optimize sampling of the posterior, when using NUTS/HMC?

If not, is there something I can do to specify the conditional independencies to make the sampling faster?

2

Welcome to the Stan forum!

To the best of my knowledge Stan does not (try to) detect conditional independencies to optimize sampling.

However, I also don’t know how excatly this optimization would work. Could you point to an article where this is done for another sampler so that people get a better idea of what you are looking after.