Hi, I had made a simple linear regression model that predicts the mean response ‘mu’ and the observed score with mean mu and standard deviation sigma, as follows: This model works fine without error with pm.Model() as batting_score_model: batting_score_model.add_coord('observations', np.arange…

Does it mean that some of my sigma estimates end up being negative and therefore throwing an error? Yes, that sounds like the problem. What about X_data, is it always strictly positive? In general I would recommend using a link function to keep a model on sigma positive, rather than specifying …

Sampling error y inf when adding sigma heteroskedastic modelling

Questions v5

jessegrabowski November 15, 2024, 12:55pm 10

Yeah great point. I’m not sure why softplus isn’t a more popular link function

Topic		Replies	Views
A little change to the sample code of MAP, however Questions	10	1136	April 1, 2018
Setting boundaries for prior distributions Questions	8	1324	February 18, 2022
ZeroDivisionError: Questions	5	2740	April 29, 2018
Need help with simple asymptote model using toy data Questions	3	454	January 28, 2021
Modeling Sampling Error Questions	17	1593	December 7, 2017

Sampling error y inf when adding sigma heteroskedastic modelling

Related topics