Running this model with Pymc4 leads to an initialization error. It seems that large values from LogNormal overflow to infinity. If I change LogNormal to smaller values like ~1-5, then it works. Interestingly,running the same model with Stan works fine. Any idea why this is happening? Is this a over…

Initialization fails when using LogNormal

ricardoV94 April 9, 2022, 7:20am 5

I am not sure if Stan is parametrizing the same way. It looks like their loc and scale may be on the output space and not on the log space like pymc

Edit: No, it seems like it’s the same scale

Topic		Replies	Views
Bad initial energy: inf or nan Questions	1	1528	March 30, 2018
Convergence issues with LogNormal v3	2	546	March 16, 2022
ValueError: Bad initial energy: inf. The model might be misspecified error replicating Multilevel Regression and Poststratification with PyMC3 notebook Questions	2	1615	May 11, 2018
Negative Binomial Distribution with logp = -inf Questions	14	1567	July 11, 2018
Runtime Exception with NUTS v3	16	1138	March 16, 2022