Seth Axen

I just had a look at this and got really confused. So I think this needs someone who's thought more about derivatives of dimensional quantities to take care of this...

There's also a plan to have ArviZ's plots in Plots.jl or Makie.jl. https://github.com/arviz-devs/ArviZ.jl/issues/108. My current thinking is that the steps are 1. Split diagnostics and statistics from ArviZ into smaller,...

FWIW, the `hcat` solution does not always work when the chains are generated with Turing, because the iterations field filled by Turing and the one automatically generated by `Chains` may...

Specifically, these are ECDF difference plots. Plain ECDF plots are already provided by StatsPlots and are compatible with `Chains` using the `@df` macro.

Good question, I think all args that are part of the same statement would be indented. @mateuszbaran, do you have any thoughts?

> OK - I'm not at all familiar with Zygote. Would supporting it mean (a) adding Zygote as a dependency and (b) adding some new methods to some Zygote functions...

We can also add utilities to aid in implementing custom distributions. As an example, consider `Restricted{Sphere,Distributions.MvNormalCanon}`. In general, all we can do with this distribution is write down enough of...

In general you can exactly sample through rejection sampling. However, if the submanifold looks like a shell like a sphere, this will essentially never produce a valid sample. This works...

My intuition says that more generally if a constraint that defines a submanifold can be written with an equal sign, then rejection sampling is not applicable.

> Right, but almost all manifolds we have are of measure zero in the ambient space. It wouldn't be a particularly useful distribution type here. Yes, that's true for our...