I have a hierarchical model of homogeneous Poisson processes where I have many groups (3000+) of event sequences, which I assume come from the same population. I model the group Poisson rates using a simple population hierarchy with non-centered parametrisation: coords = {"obs": data.index.values, …

[image] Genesis: Why did clipping cause issues in this particular case, considering that no samples were actually clipped? Divergent samples don’t get accepted so you won’t see them. Clipping causes the space below zero to be flat which nuts cant handle. The best is to soft clip like this fun…

Diagnosing divergences when clipping is involved

Questions version agnostic

Genesis February 5, 2024, 8:17pm 11

That’s a great suggestion! That’s for sure helpful with highly correlated variables; I guess it’s less likely to identify funnels, but still a great start!

Topic		Replies	Views
Advice with hierarchical model Questions	9	2258	October 8, 2018
Questions about the convergence of sampling the Mixture of hierarchical model Questions	2	1009	November 9, 2018
General doubt about problems when sampling a model Questions	7	713	March 27, 2021
Why doesn't this pymc3 model show shrinkage? Questions	6	1051	March 23, 2020
Hierarchical Binomial Model - Not converging Questions	2	906	February 1, 2019

Diagnosing divergences when clipping is involved

Related topics