ben18785

Here's a Heath Robinson version of the poll. Please edit the below to add your name next to the answer that best represents your response: - Interfaces with other packages...

Hmm, interesting. Would it be an issue though that a log-likelihood contains data, so a user might think it's somehow sampling from a sorta posterior distribution? (Whereas I guess it...

Ha, I'm glad we recently changed pints.GaussianLogPrior(mu, tau) in #1256 -- before this was failing silently! I will take a look. Thanks.

This is actually an interesting one: more interesting than it might first appear! It also impinges on how we'd want hierarchical models to work. In a non-hierarchical model, we'd probably...

Going to add a return -log infinity if sd < 0 in the eight schools model only.

I've actually realised that Stan doesn't save a point-wise log-likelihood as it runs. Instead, it computes it afterwards using each posterior sample. I think, however, that we should probably try...

Looks really good Simon -- thanks! On Wed, Jun 3, 2020 at 2:49 PM Simon Marchant wrote: > I'm looking at doing a hierarchical prior function, to at least add...

Hi David. Did you use parameter transforms for NUTS? We’ve found that NUTS runs much faster when using them (which is what Stan does)... > On 8 Mar 2021, at...

Interesting. I’m not sure what your priors are but if they’re uniform you could try the boundary transformation? Do you know how many steps the NUTS algorithm is taking? >...

Yep, the boundary at zero should be fine. When you say “slower”, it’d be good to know if this was in terms of effective samples per second and/or ESS per...