Bayesian Regression — The Hogg Way

For regression where you need proper uncertainty, outlier robustness, or heteroscedastic error modeling, use Bayesian regression with PyMC. This skill follows the approach from Hogg, Bovy & Lang (2010): start with the simplest model, diagnose where it fails, and upgrade the likelihood to match the data's actual noise structure.

The existing regression bundle uses XGBoost for nonlinear tabular prediction. This bundle is for when you need interpretable coefficients with honest uncertainty — the kind of regression that goes in a paper, a regulatory filing, or a causal analysis.