bambinos/bambi

What it solves

Bambi simplifies the process of building and fitting Bayesian linear models in Python. It specifically targets the complexity of fitting mixed-effects models, which are common in social science research, by providing a high-level interface that removes the need to write complex probabilistic programming code from scratch.

How it works

Bambi acts as a high-level wrapper around the PyMC probabilistic programming framework. It allows users to define models using a formula-based syntax (e.g., Reaction ~ Days) and handles the underlying PyMC model construction, prior selection, and sampling. It integrates with ArviZ for model diagnostics and visualization of results.

Who it’s for

Researchers and data scientists, particularly those in the social sciences, who need to perform Bayesian inference on linear or logistic regression models without needing to be experts in probabilistic programming.

Highlights

• High-level interface for PyMC, making Bayesian model building easier.
• Support for mixed-effects models common in social sciences.
• Formula-based model specification for quick iteration.
• Integration with ArviZ for summary statistics and trace plots.