Project for "Machine Learning for Data Cleaning", Winter 2019, University of Waterloo
In this project, we use methods of statistical causal inference to perform hypothesis tests of assumed causes for data errors.
We use existing data sets as linked in the Python notebooks.
Given the data set, we reformat it such that it can be loaded in Data X-Ray. For this, use the create-*.ipynb notebooks.
Once we have an assumed cause, we perform the actual causal analysis. These analyses can be found in the causal-analysis-*.ipynb notebooks.
Commonly used code for propensity score stratification is in utils/treatment_effect.py, code to read X-Ray output in utils/xray_util.py