Simpsons paradox and mixed models - large number of groups

[cluhmann]

The arviz documentation is pretty good about explaining the various diagnostics. Here is the page on effective sample size.