A practical guide to building, checking and improving Bayesian models.
The Bayesian workflow is the iterative craft of building, fitting, checking and improving Bayesian models. You write down a generative model of your data, fit it with tools like PyMC or Stan, run diagnostics, compare models, and refine — looping back as often as needed. A failed posterior predictive check isn’t a dead end; it’s the workflow telling you exactly where to look next.
This page collects the best Learning Bayesian Statistics resources on the topic: the flagship episode with the authors of the Bayesian Workflow book, more episodes going deep on workflow habits, and hands-on guides — including the open-source baygent skills that teach AI agents to follow a rigorous Bayesian modeling workflow.
With the authors of the Bayesian Workflow book
Andrew Gelman, Aki Vehtari and Richard McElreath walk through the modern Bayesian workflow: prior predictive checks, reverse Bayes, assumption dominance, and how to make modeling decisions that hold up in the real world.
Listen to the episode →MCMC re-runs from scratch on every new dataset. Amortized Bayesian inference pays the cost once -- then gives you a posterior in m…
Read the guide →Coding agents confidently say "X causes Y" without drawing a DAG, checking assumptions, or running refutation tests. I built an Ag…
Read the guide →A researcher stress-tested my Bayesian Agent Skill on a real hierarchical model. It caught the workflow violations but missed a su…
Read the guide →The Bayesian workflow is the iterative process of building, checking and improving a Bayesian model: you specify a generative model, fit it, run diagnostics, compare models, and refine — until the model captures the data-generating process well enough for your purpose.
A pipeline is a fixed sequence of steps; a workflow is iterative. In the Bayesian workflow you constantly loop back: a failed posterior predictive check or a divergent transition sends you back to rethink the model, the priors, or even the data — that feedback loop is the point.
Start with the fundamentals of Bayesian modeling, then listen to the featured Bayesian Workflow episode above with Andrew Gelman, Aki Vehtari and Richard McElreath, and practice with PyMC and ArviZ — their diagnostics (R-hat, ESS, divergences) make the workflow concrete.
PyMC and Stan for fitting, ArviZ for diagnostics and model comparison, and the Intuitive Bayes courses for hands-on practice. The baygent skills on this site also teach AI agents to follow a rigorous Bayesian modeling workflow.
Want to practice? Explore all episodes or dive into the blog for more applied Bayesian modeling.