Hi all, I am new to stan and have a basic question. I want to revise the Metropolis-within-Gibbs sampler to something like the NUTs-within-Gibbs sampler by Rstan. Basically, we have a big Gibbs sampler to sample each parameter block on each iteration. For a few parameters whose priors are not conj…

Once every year or so, I think about this. I’ve never done anything serious about it though. Suppose \theta are the parameters for NUTS and \phi are those for RWMH/Gibbs. No, Stan can’t do one or more proposals from P(\theta | \phi, X) and then pass this to some external program while holding its in…

You may want to check out Turing, which is written in a high-level programming language (Julia) and would be much more amenable to this. It already has Gibbs sampling, and adding NUTS-within-Gibbs shouldn’t be all that hard if you want to make a pull request.

NUTS-within-Gibbs by stan

General

maxbiostat November 4, 2022, 1:28pm 3

You might want to look into this thread, by @s.maskell.

Topic		Replies	Views
Using Stan as part of a Gibbs sampling algorithm General	1	575	January 13, 2024
STAN call within Gibbs sampler Modeling	1	580	April 4, 2022
Generated quantites as a mechanism to define model-specific proposals Developers	3	564	June 5, 2020
Using Stan HMC as Metropolis-Within-Gibbs step in C++ Interfaces	2	888	March 22, 2020
Two step-gibbs sampling algorithm vs block gibbs sampling algorithm Modeling	1	184	July 15, 2024

NUTS-within-Gibbs by stan

Related topics