That makes a lot of sense, being able to reason about your group mean and the level of shrinkage seems like a clear win for that parameterization. With that parameterization, the models issues a few warnings:
Auto-assigning NUTS sampler...
Initializing NUTS using jitter+adapt_diag...
Multiprocess sampling (2 chains in 2 jobs)
NUTS: [a_cluster, sigma, a]
Sampling 2 chains, 9 divergences: 100%|██████████| 5000/5000 [00:02<00:00, 2024.95draws/s]
There were 2 divergences after tuning. Increase `target_accept` or reparameterize.
There were 7 divergences after tuning. Increase `target_accept` or reparameterize.
What changes if any would you make with these warnings? Feel free to link to some more great resources if anything comes to mind.
I actually already ordered Rethinking(second edition), I’ll keep an eye out for your PyMC port and I’ll definitely check out your podcast 
. Thanks for you help!