This repository hosts all the code related to my master seminar thesis 'Causal inference through the front-door' at the chair of Marketing of Prof. Dr. Dominik Papies at the Faculty of Economics and Social Sciences of the University of Tübingen. Generally this repository is divided into two parts:
- 01_simulations: hosts R scripts that challenge the FDC using simulated data
- DAG_Simulations.html: provides a theoretical and empirical introduction into the Front-Door Criterion (FDC).
- DAG_Simulations_Confoundedness.html: examines violations of the FDCs identifying assumptions using simulated data.
- DAG_Simulations_BinaryVariables.html: provides empirical evidence on causal identification of the FDC using simulated data for binary treatment, mediator and outcome variables.
- Final_Thesis_Results.R: R-Script that runs the simulations that are finally addressed in the seminar thesis itself.
- Final_Presentation_Results.R: R-Script that runs the simulations that are finally addressed in the presentation of my seminar thesis.
- 02_taxi: hosts R scripts for the replication of the empirical application of the FDC by Bellemare et al. (2020)
- 01_local: contains the initial R-scripts that were used for data import and first estimation steps + examines the results that were computed on the bwUniCluster and stored in a csv-file.
- 02_bwUnicluster: contains the R as well as the Rout files from the bwUniCluster. Here, the steps of data import, preparation and final estimation were done in separate steps and scripts.
References:
Bellemare, M. F., Bloem, J. R., and Wexler, N. (2020). The paper of How: Estimating treatment effects using the front-door criterion. Working paper.