I would probably parametrize it in terms of a lognormal or something that has a concept of mean and dispersion. HalfNormal is limited. For instance, you can’t ever learn that most sigmas are tightly grouped around 3 (or any other non-zero value)
I would probably parametrize it in terms of a lognormal or something that has a concept of mean and dispersion. HalfNormal is limited. For instance, you can’t ever learn that most sigmas are tightly grouped around 3 (or any other non-zero value)