Sort this notes

[aloctavodia]

These notes were in the readme, those still useful should be somewhere else (maybe open issues for some of them).

Features, warnings, and details

• You can start and stop the simulations This is pretty robust, and works the way you would expect even if you have set a random seed. This is useful if you want to peek at the histogram early.
• The progressbar is extra informative Notice that it collects all the warnings from each PyMC sampling run and aggregates it for you. If you are getting tons of divergences, maybe you do not need simulation based calibration to know your model has some problems?
• This takes a long time It is embarrassingly parallel, but this implementation will not help you with that.
• Other rank statistics You can add a pm.Deterministic variable to your model to compute other rank statistics.

Examples from the paper

TODO

• Add thinning
• Add ADVI option
• Run on all examples from paper