BinomialBias

This library computes and plots quantitative assessments of discrimination within organizations, based on the binomial distribution.

This code supports the following paper:

Quantitative measures of discrimination with application to appointment processes. Robinson PA, Kerr CC (2024). PLoS ONE 19(3): e0299870. https://doi.org/10.1371/journal.pone.0299870

There are several ways to use this library, described below.

Webapp

A live webapp is running at https://binomialbias.sciris.org.

Local installation and usage

Python

To use locally with Python, run

pip install binomialbias

This can then be run via e.g.:

import binomialbias as bb
bb.plot_bias(n=20, n_e=10, n_a=7)

This example shows the statistics for the case where there were n = 20 appointments (e.g., the size of a committee), out of which n_e = 10 people were expected to belong to a given group (e.g., female), and for which n_a = 7 actually were.

Shiny

To run the Shiny app locally, clone the repository from GitHub, then install with

pip install -e .[app]

The Shiny app can then be run locally via the run script.

Structure

• All code for the Python package is in the binomialbias folder.
• The script for generating the figure in the paper is in the scripts folder.
• Continuous integration tests are in the tests folder.
• Older Jupyter and Matplotlib versions are available in the archive folder.