Bayesian Methodologies with pyhf

TL;DR

This paper introduces bayesian_pyhf, a Python package that integrates pyhf and PyMC to enable parallel Bayesian and frequentist analysis of complex statistical models, enhancing flexibility in high-energy physics data analysis.

Contribution

The paper presents bayesian_pyhf, a novel Python package that combines pyhf and PyMC for efficient Bayesian inference on models built with HistFactory, supporting parallel analysis methods.

Findings

Enables parallel Bayesian and frequentist evaluations.

Supports complex multi-channel binned models.

Facilitates Bayesian analysis with existing high-energy physics models.

Abstract

bayesian_pyhf is a Python package that allows for the parallel Bayesian and frequentist evaluation of multi-channel binned statistical models. The Python library pyhf is used to build such models according to the HistFactory framework and already includes many frequentist inference methodologies. The pyhf-built models are then used as data-generating model for Bayesian inference and evaluated with the Python library PyMC. Based on Monte Carlo Chain Methods, PyMC allows for Bayesian modelling and together with the arviz library offers a wide range of Bayesian analysis tools.