There are some problems about the inverse Gaussian process. When I estimate the parameters of an oil leakage process as an inverse Gaussian process, that is, the leakage increment obeys the inverse Gaussian distribution, with pm.Model() as model: w = pm.Uniform('w', lower=0, upper=100) k = …

Looks like you are hitting numerical precision issues, possibly due to an extreme combination of data and/or parameters. Can you narrow down which value(s) from your observations leads to an underflow of -inf for the likelihood?

There are some problems about the inverse Gaussian (Wald distribution)

Questions

J.Cui June 22, 2021, 3:54am 7

Thank you！ I found some problems with my data. After changing the data, there is no -inf

Topic		Replies	Views
Truncated Inverse normal distribution (also known as Wald distribution) Questions	15	2954	April 27, 2018
Help understanding why sampling fails in Zero-Inflated Gamma likelihood model Questions	4	2515	November 29, 2017
Use of Inverse Gamma for standard deviation and overflow ("ValueError: Mass matrix contains zeros on the diagonal") Questions	2	593	February 21, 2019
Gamma GLM problem: ValueError: Mass matrix contains zeros on the diagonal Questions	3	753	November 30, 2018
Inference from binary data & proper way to use GaussianRandomWalk Questions	2	803	September 15, 2017

There are some problems about the inverse Gaussian (Wald distribution)

Related topics