Rolling Regression in PyMC

Yeah you can definitely get creative with the priors. Then the regressions aren’t independent anymore, but this should be seen as an advantage of the Bayesian method vs vanilla OLS (why should they be independent?)