Pairwise share ratio interpretations of compositional regression models

This repository contains all codes necessary to reproduce the results of our article “Pairwise share ratio interpretations of compositional regression models”, available at:

Dargel, Lukas, and Christine Thomas-Agnan (2024). Pairwise share ratio interpretations of compositional regression models. Computational Statistics & Data Analysis, Volume 195, July 2024, 107945 Accessed 08 Mar 2024. Online at https://doi.org/10.1016/j.csda.2024.107945

The vignette accompanying the article illustrates the presented mathematical tools for a case study of French electoral sociology and can be downloaded directly as HTML.

Notes on reproducibility

To reproduce the results, you can run the bash commands below:

# get all codes from GitHub
git clone https://github.com/LukeCe/paper_Share_Ratio
cd paper_Share_Ratio # change working directory to the cloned git

# build the docker image for the paper from the current directory
docker build -t repro_share_ratio . 

# run the analysis from within the docker container
docker run --rm \
  -v .:/home/paper_Share_Ratio \  # mount the repository into the docker
  -w '/home/paper_Share_Ratio' \  # change working directory accordingly
  repro_share_ratio \             # specfiy the docker image
  Rscript repro.R                 # execute the repro.R script

Executing the command Rscript repro.R performs the following steps:

1. Download the three source data sets and places them in in\data:

   • Election results from the French Ministry of the Interior
   • Socio-demographic information from the population census provided by INSEE
   • Geographic contours of French municipalities provided by IGN (this file has about 250MB and can take some time to download)

2. Merge the data sources and place the combined data in out/data

3. Rebuild the HTML vignette that generates all results used in the article

   • The vignette is placed in notebooks
   • The outputs used for the article are placed in out/figures and out/tables

How to cite

Please cite the above article when using any material from this repository.

Licenses

Text and figures : CC BY 4.0

Code : GPL 3.0

Data : See the policies of the original data providers