Using a different MCMC sampler?

[nluetts]

Hi 👋

I just read your article over at ChemRxiv and found it pretty interesting.
I noticed that for MCMC sampling you use the Metropolis-Hastings algorithm and need a lot of samples, and I was wondering:
why not use a more efficient sampler such as Gibbs or NUTS?
Without making a deep dive into your data evaluation (kudos that you put it on Github, by the way), this may be a naive question, but just in case you have not yet seen this:
Turing.jl maintains a package for Julia which implements NUTS, https://github.com/TuringLang/AdvancedHMC.jl.

Feel free to close this issue at any time :-)

Regards
Nils