Using machine learning and satellite imagery to estimate aid's effect on wealth: comparing China and World Bank programs in Africa

Code in support of a thesis for a Masters of Computational Social Science at Linköping University.

• Thesis and Appendices (PDF)

• AnalysisSteps explains the order the files were executed.

Overview:

• R: contains majority of the analysis code. The code
  • writes .csv files to the ./data/interim directory which are read by later scripts.
  • writes maps, charts, and figures to a ./figures directory
  • writes tables to a ./tables directory
  • reads shapefiles, aid project, and confounder data from a ./data directory
  • writes results to a ./results directory
• python: scripts used to download satellite imagery over DHS points to NAISS (National Academic Infrastructure for Supercomputing in Sweden)
• scripts:
  • batch utility scripts to run on Windows laptop and NAISS environment
  • slurm scripts to exeucte long-running R scripts on NAISS
• env_conf: apptainer recipes to configure the NAISS python and RStudio environment where analysis runs