I am playing around with the great example in Simpson’s paradox and mixed models — PyMC example gallery and I noticed the following: i) Increasing the number of groups from 5 to 50 makes it necessary to use pm.sample(target_accept=0.999, tune=2000, draws=10000) in order to not get any warnings ab…

Simpsons paradox and mixed models - large number of groups

ricardoV94 March 7, 2024, 3:57pm 4

The original model is probably already bad

Topic		Replies	Views
Sampler occasionally gets static in pymc3.1, but not in pymc3.0 Questions development	19	2271	July 5, 2018
Improve Sampling of Multilevel Regressions Model Questions	13	1270	March 9, 2020
Multinomial hierarchical regression with multiple observations per group ("Bad energy issue") Questions	22	3450	March 14, 2019
Problems with Hierachical Bayesian Model - Advertising Questions	16	1081	August 10, 2018
Hierarchical linear model with two parent distributions Questions	16	1338	September 15, 2018