Greetings all, I’m struggling with the fluctuating Rhat values. As far as I know, the default value for the tune is 1000, yet the problem is, the higher the tune value, the higher Rhat outcome (particularly with random effects) I get or vice versa. Is this a problem or am I missing something? The v…

Sure thing! Since the reason is overdispersion, I’ve used NegativeBinomial rather than Poisson. So I’ve added the prior for the alpha parameter, # Prior for the dispersion parameter alpha = pm.Gamma('alpha', alpha=2, beta=0.1) and then run pm.NegativeBinomial() obs = pm.NegativeBinomial('ob…

How does 'tune' parameter impact Rhat values?

Questions v5

fatihbozdag December 11, 2023, 4:55pm 2

Well, it seems like the issue was due to overdispersion. The issue is now solved.

1 Like

Topic		Replies	Views
Rhat Nan Metropolis Questions	2	1062	April 7, 2018
Bad r-hat on intercept only v5 modeling	7	1147	August 18, 2022
Sampler occasionally gets static in pymc3.1, but not in pymc3.0 Questions development	19	2271	July 5, 2018
Hierarchical model where convergence is difficult: chains look bad. Any tips? Questions	6	1051	January 10, 2019
Hierarchical linear model with two parent distributions Questions	16	1338	September 15, 2018

How does 'tune' parameter impact Rhat values?

Related topics