@tboutelier
Hello,
I am discovering bayesian analysis and I read your post with interest. I plan to write a custom likelihood function. I have not exactly the same likelihood as yours but it shares some similarities. And, I wonder why, in your code, the logp return is -N / 2 * (yty - ytf ** 2 / ftf) and not -N / 2 * T.log(yty - ytf ** 2 / ftf)
Best regards
Stephane
1 Like