Hi, I am looking at some RNA count data where the count data is modelled as a negative binomial distribution. The problem however is that the logp of the model is -inf. I read an earlier thread where another user had the same issue, and the problem were nan values in the input. However, in my case …

Negative binomial returning -inf logp

ckrapu July 15, 2020, 3:22pm 9

I’m glad to hear you solved it. I often use Theano with the 32 bit floating point representation so I will look out for this issue in the future.

Topic		Replies	Views
Negative Binomial Distribution with logp = -inf Questions	14	1567	July 11, 2018
Log probability negative infinity Questions	4	1323	December 31, 2019
Get nan or inf from model.logp (model.test point) is an attestation of incorrectly configured model? Questions	22	8333	June 24, 2017
Average Loss = inf Questions	7	2511	March 21, 2018
"Some of the observed values are associated with a non-finite logp" v5	2	729	September 16, 2023