alterlab-pymc

To build and sample hierarchical Bayesian models in PyMC, you prepare your dataset, construct the model structure, execute sampling, and run diagnostics using ArviZ to validate posterior estimates and posterior predictive checks.

The standard Bayesian modeling workflow in PyMC covers data preparation, model construction for linear or hierarchical examples, sampling, convergence diagnostics, posterior predictive checks, and final predictions.

Yes, you can compare multiple Bayesian models using LOO and WAIC. The workflow provides educational examples and reference patterns for model comparison, allowing you to evaluate linear versus hierarchical models and select the best fit.

Yes, this Bayesian workflow supports posterior predictive checks and predictions. It includes templates and practical use cases to validate probabilistic models and generate predictions from the posterior distribution using PyMC and ArviZ.

For Bayesian modeling and diagnostics with PyMC, you need ArviZ for posterior analysis, NumPy and Pandas for data preparation, and Matplotlib for visualizing posterior predictive checks and model diagnostics.

You should use hierarchical models instead of linear models in PyMC when your data has nested or grouped structures. The workflow includes ready-to-use templates for both linear and hierarchical models to guide your model construction.