Bayesian Analysis with pyhf using PyMC

The aim of this project is to provide tools for Bayesian analysis of pyhf models using the Python library PyMC.

Setup

To set up a Python virtual environment, download the following dependency file environment.yml and run:

conda env create --file environment.yml

You can then run

pip install --editable .

for an editable install of bayesian_pyhf.

For a quick example, see example.ipynb.

Development

Lock files

To create a fully reproducible environment lock file from the high level environment.yml, install conda-lock and then create a hash-level conda-lock.yml lock file with

conda-lock lock --file environment.yml --kind lock

or more simply, with nox installed, by running the nox default session (which requires Docker)

nox

An environment can then be created from the lock file either with

conda-lock install --name bayesian_pyhf conda-lock.yml

or

conda env create --file conda-lock.yml

See conda-lock install --help for additional options.

Updating the environment

To add new dependencies to the environment definition files simply add the dependencies to the environment.yml and then rebuild the lock file. To update your existing environment from an updated environment.yml file use

conda env update --name bayesian_pyhf --file environment.yml