Chapter 2. Programming Probabilistically – A PyMC3 Primer

Now that we have a basic understanding of Bayesian statistics we are going to learn how to build probabilistic models using computational tools; specifically we are going to learn about probabilistic programming. The main idea is that we are going to use code to describe our models and make inferences from them. It is not that we are too lazy to learn the mathematical way, nor are we elitist hardcore hackers—I-dream-in-code. One important reason behind this choice is that many models do not lead to a closed-form analytic posterior, that is, we can only compute those posteriors using numerical techniques. Another reason to learn probabilistic programing is that modern Bayesian statistics is done mainly by writing code, and since we already know Python, why would we do it in another way?! Probabilistic programming offers an effective way to build complex models and allows us to focus more on model design,...