Metadata-Version: 2.1
Name: popodds
Version: 0.0.5
Summary: Simple package for Bayesian model comparison.
Home-page: https://github.com/mdmould/popodds
Download-URL: https://github.com/mdmould/popodds/archive/refs/tags/0.0.5.tar.gz
Author: Matthew Mould
Author-email: mattdmould@gmail.com
License: MIT
Requires-Python: >=3.7
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy
Requires-Dist: scipy
Requires-Dist: kalepy

# popodds
Simple package for Bayesian model comparison.

Given samples from a posterior distribution inferred under some default prior, compute the Bayes factor or odds in favour of a new prior model.

## Usage

The package consists of the `ModelComparison` class to compute Bayes factors, and a wrapper function `log_odds` for simplicity.

The computation only requires a few ingredients:
- `model` a new prior model, e.g., samples from a simulation,
- `pe_samples` samples from a Bayesian parameter estimation run,
- `pe_prior` a function, prior evaluations, or prior samples corresponding to the original parameter estimation prior,
- `model_bounds` optional parameter bounds for the new prior model,
- `pe_bounds` optional parameter bounds for the original prior model,
- `prior_odds` optional odds between the prior models, which defaults to unity.
