How can I effectively propagate parameter uncertainties from one hierarchical level to the next?

Questions version agnostic
,
6

There’s also a multivariate normal approximation: prior_from_idata — pymc_experimental 0.0.15 documentation