Pakistani_Audits_Project

Repository for the project on optimizing audit schemes with machine learning.

There are currently two main folders in this repo which house code and data: \build and \analysis.

The \build folder contains all data and code necessary to clean and pre-process the data for analysis. It contains several sub-directories:

1. \input which houses the raw data on (XYZ audit data). 1.1. NOTE: if any of the data files are > 100MB, we can't store them on Github directly. In that case we'll have to add the data files to the .gitignore file so that when we commit the code updates, the commit doesn't try to push the data files themselves. We'll also have to add the output datasets to .gitignore also. That means that we'll be storing the data locally or on Dropbox and having code stored on Github (and lcoally) point to it, but we won't store the data itself on Github.
2. \code which houses the cleaning and pre-processing script(s).
3. \output will contain the cleaned dataset(s) which are imported directly from the scripts in the \analysis folder.
4. \temp contains any log files or temporary outputs which we may want to save.

The \analysis folder contains all code to perform econometric analysis and create tables and figures for the paper.

1. \input will contain the link to the data in \build\output. Note that any pickled ML algorithms produced by the scripts in \analysis\code will also be saved here to be used as inputs in
2. \code contains the analysis scripts. There are two types of scripts: (1) scripts producing necessary ML algorithms and (2) scripts producing LaTeX tables, figures, etc. ML algorithms are compute-intensive, so it's best practice to create those once, then save the algorithms themselves in the \inputs folder to be called if necessary in the analysis scripts that make tables.
3. \output contains LaTeX tables and other figures/plots, etc.
4. \temp contains any log files or temporary outputs which we may want to save

Script Order:

1. The ~\build\code\grf_pre_cleaning.R script imports the raw data files and processes them for ML exercises
2. The ~\analysis\code\GRF_train_main.R script trains and saves the ML algorithms
3. The ~\analysis\code\optimization_testing_script.R (which will eventually be changed to an official optimization script) performs the MVPF minimization exercises

*NOTE: the .gitkeep files are placeholder files that are there so Git tracks the otherwise empty folders. Once we add contents to a folder, we can remove the .gitkeep file