Prediction Gaps: Rethinking educational outcomes through differences in model performance

This repository contains the code and scripts used to generate the figures in the paper "Prediction Gaps: Rethinking educational outcomes through differences in model performance" by Javier Garcia-Bernardo, Weverthon Barbosa Machado, Eva Jaspers, Samuel Plach4, and Erik Jan van Leeuwen.

Repository Structure

• Code_figures/
  • 5_paper_figures.ipynb: Main Jupyter notebook for generating all figures and tables for the paper.
  • config.py: Configuration file with paths, plotting styles, and variable descriptions. For replication, change the paths here to point to your local directories.
  • helper.py: Custom helper functions for data processing and visualization (not shown here, but referenced in the notebook).
• CBS_export/: Directory for the scripts used at CBS and the exported data required to run the notebook.
• README.md: This file.

Requirements

• Python 3.11+
• See environment_cbs.txt for a full list of required packages at CBS.
• See environment.txt for a full list of required packages to recreate the figures

Installation and Usage

• Set up the environment

conda env create -f environment.yml

• Run the notebook

Start Jupyter and open the main notebook:

jupyter notebook Code_figures/5_paper_figures.ipynb

Make sure to adjust the paths in Code_figures/config.py to match your local directories before running the notebook.

Citation

If you use this code or figures, please cite the associated paper (citation to be added).