How to use the posterior distribution of one model as a prior distribution for another model

[perrette]

I think the question is now “solved” for the common cases of Normal and LogNormal distributions. The more general approach with any kind of parameterizable distribution is one step away, by using the intermediate step of transforming to a uniform distribution. I don’t need it for my work so I leave it to others to implement. Here it is:

How best to use a posterior sample as a prior for a future integration in another model?

FYI, here is an implementation that I tested and which seems to work. It is designed to deal with Normal and LogNormal distributions, which should already cover many practical cases. import numpy as np from scipy.stats import norm, lognorm import pymc as pm import arviz def define_params_from_trace(trace, names, label, log_params=[]): """Define prior parameters from existing trace (must be called from pymc.Model context) Parameters ---------- trace: pymc.InferenceData fro…