How to decide on what priors distributions to use for parameters?

Mmmh ok, I think this NB could help you go in the right direction then. I’ve never done this personally though

Thank for the link, I think I got rough idea. Another question, if I want to design custom distribution, for instance, noncentral t like in scipy nct. What do I need in order to create a custom distribution with pymc3 and choose priors for it?