In the original code, the priors of alpha_0 and alpha_1 are given by normal distributions. When 9f0sdau90 updates the priors, s/he fits a gaussian kernel to the posterior samples. Given that, I’m just curious why you chose to use a student t distribution for the updated priors instead of a normal distribution, like (eg) this:
alpha_0 = pm.Normal(‘alpha_0’, mu=mu0, sd=sd0)
alpha_1 = pm.Normal(‘alpha_1’, mu=mu1, sd=sd1)