Hierarchical Logistic Regression with Binomial Likelihood - The chain reached the maximum tree depth, The rhat statistic is larger than 1.05 for some parameters, The estimated number of effective samples is smaller than 200 for some parameters

vanko · December 16, 2021, 6:45pm

Thank you @cluhmann for taking look into this.
Sorry for taking time to get back to you. I started run it on GPU system and had to Install Linux with Ubuntu, set conda and pymc3 env on it to preform better speed.
So my latest modification on this Hierarchical model based on your recommendation is reducing target_accept and increase tune to 8 000, also draws=10000:

trace_hr_nc = pm.sample(chains=2,
draws=10000,
tune=8000,
# max_treedepth = 15,
return_inferencedata=True)

As result - tons of divergences and following :

Topic		Replies	Views
Hierarchical Logistic Regression. How to reparameterize? Questions	1	456	October 26, 2021
Hierarchical Binomial Model - Not converging Questions	2	906	February 1, 2019
Hierarchical Linear Regression Model -- why is it not converging? Questions	3	1586	March 14, 2018
Hierarchical Binomial regression - ineffective sampling Questions	4	922	January 17, 2018
Hierarchical Logistic Regression sampling problems Questions	4	1828	January 18, 2018

Hierarchical Logistic Regression with Binomial Likelihood - The chain reached the maximum tree depth, The rhat statistic is larger than 1.05 for some parameters, The estimated number of effective samples is smaller than 200 for some parameters

Related topics