Seth Axen
Seth Axen
No, `loo` is exported by PosteriorStats. But actually, on second thought there'd be no benefit in overloading `loo` for `Chains`, since its input is an array of log-likelihood values that...
Agreed this should be fixed upstream. @torfjelde can you by any chance share an MWE?
Thanks @torfjelde , that should be enough for me to reproduce the issue when I get to a computer.
@torfjelde I just tried and wasn't able to reproduce this. Could you write your chains to a file and share them with me?
@torfjelde gentle reminder to send the problematic chains when you get the chance
@torfjelde _slightly_ less gentle reminder to send the problematic chains
I'm not certain what the original issue is, but if it's the same as #447, then MCMCDiagnosticTools v0.3.10 contains a fix for this: https://github.com/TuringLang/MCMCDiagnosticTools.jl/pull/119
> Okay, so combining #246 and [TuringLang/DynamicPPL.jl#462](https://github.com/TuringLang/DynamicPPL.jl/pull/462), this now works: Awesome! Can you also check the estimated metric? > Though I'm confused as to why `MCMCChains` reports a non-NaN ESS...
> > If they're even a little offset from 1, then a NaN won't be returned, e.g. > > Yeah figured it had something to do with that, but was...
More generally, `bijector(d)` for a given continuous univariate distribution `d` is guaranteed to be strictly monotone, so we have 2 cases: the bijector is increasing or decreasing. Suppose we have...