Suppose there are three independent normal variables and their means are 0 and their standard deviations have independent logarithmic priors. Suppose we have one observation for each, 1.0, 1e6 and 1e-4. import pymc as pm def abs_log_minus(tensor): return -pm.math.log(pm.math.abs(tensor)) wi…

You logp definition allows for negative values, which when used as a scale yield a null likelihood. This is exactly what happened: Starting values: {'std': array([ 0.17563482, 0.95619595, -0.09898284])} If you play with the seed you will probably also get the model to fail at first with 2 datapoi…

pymc.exceptions.SamplingError for three normal variables the means whereof are 0 and the standard deviations whereof have independent logarithmic priors

Questions v5

JDawson-Camlin March 16, 2023, 1:54pm 3

Thank you, @ricardoV94.

1 Like

Topic		Replies	Views
PyMC3 does not sample from my custom prior Questions	1	826	January 4, 2019
Sampling with a general lognormal model Questions	9	1004	December 3, 2019
How to formulate a Lognormal likelihood using PyMC3? Questions	19	2450	March 10, 2021
Trying to use a prior distribution inside a logarithm Questions	5	801	February 11, 2020
Pymc3.model likelihood evaluation for given prior sample Questions	2	550	June 22, 2018

pymc.exceptions.SamplingError for three normal variables the means whereof are 0 and the standard deviations whereof have independent logarithmic priors

Related topics