Arviz: Add R* convergence diagnostic

Created on 2 Sep 2020  路  2Comments  路  Source: arviz-devs/arviz

Add R* diagnostic as described in

  • Ben Lambert and Aki Vehtari (2020). R*: A robust MCMC convergence diagnostic with uncertainty using gradient-boosted machines. arXiv preprint arXiv:2003.07900

The reference R code available at

Most helpful comment

In case it's useful, here's an example calling MCMCChains.jl's implementation of R* on the same data: https://gist.github.com/sethaxen/b22d9a9ca70530d8f0bce1687a09633f.

All 2 comments

Adding some initial tests here

https://gist.github.com/ahartikainen/b892b487df9c067d8e91cdd4d986d3af

I'm not sure what library we should use. And if we select some, then I think it should be "optional".

I think we should also do a comparison test between different libs to see how they perform (time vs accuracy).

In case it's useful, here's an example calling MCMCChains.jl's implementation of R* on the same data: https://gist.github.com/sethaxen/b22d9a9ca70530d8f0bce1687a09633f.

Was this page helpful?
0 / 5 - 0 ratings