SBC tutorial

[janfb]

we need a detailed tutorial about SBC and including help for understanding the SBC rank plots.

E.g., it would be great to build SBC intuition by showing how SBC detects bias and dispersion in a simple linear Gaussian simulator example, like here:

[psteinb]

I am happy to help with this. But it would be cool to know where the plots came from and how they were generated.

[janfb]

I generated those plots using a simple Gaussian example and calculating SBC and coverage for different amounts of biases and dispersions.

[psteinb]

Ok, to me ... this is wonderful but hard to digest. I think such plots can be the finish of this tutorial. But let's not get ahead of ourselves. I'd first take one of the corner cases you outlined and illustrate the rank plots, then go to the rank plots summary statistics (either the posterior ranks or the KS pvalues). Then add everything up and compile such a facetgrid or matrix of plots.

Cool, question would be how to go about this?
Should this be a standalone jupyter notebook or something more like this markdown (which likely was generated from a notebook)?

[janfb]

yes, totally agree. I just posted this here for reference because I had this plot lying around in some notebook.

Yes, let's plan come up with an outline of the tutorial here and one of us makes a first draft in a notebook and then we improve from there?

How about we draft an outline together in Notion and then post is as reference here?
https://www.notion.so/SBC-Tutorial-e36c4b0a0e7d4715806e6845951c36cf

[psteinb]

Sounds like a plan.