StatisticalRethinking v4 (SR4) about statisticalrethinking.jl HOT 4 CLOSED

Steps/sequencing:

These are very significant changes to SR. Not all of this will be available the first time around.

These are my current priorities for phase one:

1. Initially I'll work primarily on

   1.1. StatisticalRethinking.jl v4,
   1.2 StatisticalRethinkingStan.jl (the project) and
   1.3 StatisticalRethinkingPlots.jl.

   Thus Stan, i.e. Stan's cmdstan executable, will remain my primary means to draw MCMC samples.

2. During the first phase it's basically a refactoring of the existing code base (mainly the introduction/completion of @require barriers and moving graphics to StatisticalRethinkingPlots.jl)

3. Adapting the code base to AxisArray.jl based chains.

4. Where possible, existing functions will be updated to accept KeyedArray chains and DataFrame based input.

5. It's simple to convert an MCMCChains.Chain object into a KeyedArray chains object. A method for this will be provided. This will possibly drop some info added e.g. by Turing.

from statisticalrethinking.jl.

If you haven't already switched to AxisArray.jl, you might want to check out DimensionalData.jl, which does essentially the same thing, but seems to be better documented and has better support. The AxisArrays/DimensionalData devs have suggested merging the two packages at some point in the future.

from statisticalrethinking.jl.

Thank you @ParadaCarleton , I will certainly take a look! I'm pretty impressed with AxisKeys.jl.

AxisArrays.jl, in my opinion, for chains at least, was a bit of a mistake. But a lot can improve in a few years!

from statisticalrethinking.jl.

In StatisticalRethinking v4 most of above ideas have been implemented (and a few more in fact). For now I'll close this issue.

from statisticalrethinking.jl.