Bambi: A simple interface for fitting Bayesian linear models in Python

TOMÁS CAPRETTO; JACOB WESTFALL; RAVIN KUMAR; MARTÍN OSVALDO A.; CAMEN PIHO; TAL YARKONI

arxiv

arxiv

Año: 2020

The popularity of Bayesian statistical methods has increased dramatically inrecent years across many research areas and industrial applications. This isthe result of a variety of methodological advances with faster and cheaperhardware as well as the development of new software tools. Here we introduce anopen source Python package named Bambi (BAyesian Model Building Interface) thatis built on top of the PyMC3 probabilistic programming framework and the ArviZpackage for exploratory analysis of Bayesian models. Bambi makes it easy tospecify complex generalized linear hierarchical models using a formula notationsimilar to those found in the popular R packages lme4, nlme, rstanarm and brms.We demonstrate Bambi´s versatility and ease of use with a few examples spanninga range of common statistical models including multiple regression, logisticregression, and mixed-effects modeling with crossed group specific effects.Additionally we discuss how automatic priors are constructed. Finally, weconclude with a discussion of our plans for the future development of Bambi.