DanielRobertNicoud

@zwelitunyiswa Unfortunately, as far as I know this doesn't work. I am not looking for sampling the posterior predictive at an unseen point $x^\star$. What I want is to compute...

I think this could be easily (?) implemented using the [`pymc.compute_log_likelihood`](https://www.pymc.io/projects/docs/en/latest/api/generated/pymc.compute_log_likelihood.html) function.

@tomicapretto Yes, that seems correct. As a simplified example, suppose I am doing a Bayesian linear regression $y = x^T\beta + \epsilon$ with $\epsilon\sim N(0,\sigma^2)$ with $\sigma$ known. Fitting the...

@tomicapretto The following works in pure `pymc` (as you can see, I test against manually computed values). ```python import pymc import pandas as pd import numpy as np import arviz...

@tomicapretto Hi, did you have time to take a look at this?

@ricardoV94 That would be a very useful addition in the case where the model has not changed! Maybe a way could be even to pass the compiled function to the...