Article: JAGS, NIMBLE, Stan: a detailed comparison among Bayesian MCMC software

[Marc Kery]

Dear all,

here's an interesting comparison between the three most widely used Bayesian modeling software pieces in Ecology, or probabilistic programming languages (PPLs), as the authors call them (although it might be better to call them dialects of a single PPL, the BUGS language):

https://arxiv.org/abs/2107.09357

One take-home message from the article is this:

"Taking into account all of our simulations, Stan appears the go-to solution if one wants to learn only one between JAGS, Stan and NIMBLE programs".

But then also this:

"The greatest limitation of Stan, as underlined by many authors, is the unfeasibility of discrete parameters, as the computation of the gradient in this case is not allowed."

But see yourself 🙂

Best regards  --- Marc

[Perry de Valpine]

Dear Marc et al,

Here is a blog post showing that, in the one of their four examples we've dug into in detail so far, the authors gave Stan a different and easier problem than they gave JAGS and NIMBLE.  When we give NIMBLE the same problem as Stan, which is easy to do, NIMBLE's default samplers perform between 2 and 45 times better than what they reported.  Each of the packages performed best in at least one of their examples.  We see issues in some of the other examples, but we haven't posted anything about them yet.  For now, please take their results with a large grain of salt.

https://r-nimble.org/a-close-look-at-some-posted-trials-of-nimble-for-accelerated-failure-time-models

-Perry

--
*** Three hierarchical modeling email lists ***
(1) unmarked: for questions specific to the R package unmarked
(2) SCR: for design and Bayesian or non-bayesian analysis of spatial capture-recapture
(3) HMecology (this list): for everything else, especially material covered in the books by Royle & Dorazio (2008), Kéry & Schaub (2012), Kéry & Royle (2016, 2020)
---
You received this message because you are subscribed to the Google Groups "hmecology: Hierarchical Modeling in Ecology" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hmecology+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hmecology/ZR0P278MB0869E65C3DBE49550D624429EB899%40ZR0P278MB0869.CHEP278.PROD.OUTLOOK.COM.

[Marc Kery]

Dear Perry,

excellent, thank you very much for pointing that out and referring to your blog post.

Best regards  --- Marc

---

From: Perry de Valpine <pdeva...@berkeley.edu>
Sent: Sunday, October 31, 2021 14:55
To: Marc Kery <marc...@vogelwarte.ch>
Cc: hmecology: Hierarchical Modeling in Ecology <hmec...@googlegroups.com>
Subject: Re: Article: JAGS, NIMBLE, Stan: a detailed comparison among Bayesian MCMC software