Cameron Pfiffer

Do you have thoughts on providing an alternative backend for nonlinear parameters? I would prefer to move that kind of thing over to a new package entirely rather than have...

We may end up having someone work on some of these issues, so I wanted to type up the two projects I think would be the most beneficial in terms...

Nah, we actually did a ton of work to remove `DataFrames` from MCMCChains internals to cut down on load times. The idea is generally that you can put whatever you...

> Namely, I think you could swap out AxisArrays with a DimensionalData.jl wrapper around a StructArray or a TupleVector, since DimensionalData is just a wrapper around any other kind of...

Woops, I missed this one as reviewable. If I forget, as I do often, you can flag me by requesting a review over here: 

Okay, thanks for adding that stuff. Looks good to me. When there's a merge conflict, it's typically the PR person's job to fix the merge conflict. Fortunately this one is...

No problem, all part of the learning process. I've just enabled tests and will check back to make sure they've passed.

From what I remember, this is a problem upstream with the GR backend for plots. Check out this issue: https://github.com/jheinen/GR.jl/issues/143. It's a bit hacky, but there is a fix outside...

> However, what about something like > > ```julia > asarray(chain, :x, (3, 2)) > ``` > > that would output a `Array{Union{Missing,Float64},4}` of size `(niterations, 3, 2, nchains)` that...

You can extract the mean column from the `ChainDataFrame` directly if everything's sorted correctly: ```julia m = mean(group(chain, var)) m.nt.mean ``` But I think you've highlighted a bigger issue, which...