Karim305 / master-thesis

Master Thesis Repository

Joint Density Modelling of Income and Wealth

This repository contains the code for the analysis in my master's thesis, titled "Joint Density Modelling of Income and Wealth", in Statistics at ETH Zurich. Since the administrative tax data used in this thesis is highly confidential, this repository contains only the code and, unfortunately, the results are not reproducible without data access.

Repository Structure

Scripts that are sourced by other scripts in order load functions, data, and common variables

• 00_utils.R: Functions used across the project.

• 01_setup.R: Setup script that loads data and creates common variables.

• /_utils: Utility scripts specific to certain methods.

  • copula.R: Utilities for copula models.
  • mbo.R: Utilities for model-based optimization.

Data preparation scripts (order matters)

• /_dataprep: Scripts for data preparation.
  • 01_dataprep.R: Initial data preparation script.
  • 02_dataprep2.R: Further preparations and scaling data.

Model fitting scripts (order usually matters within model classes)

• /_fitting: Model fitting scripts.
  • /Copula: Scripts for fitting copula models. Note that these scripts require that the regression models are fitted first. Furthermore, the copula scripts take the output (to disk) of 01_combine_models.R as input.
    • 01_fit_copula_residuals.R: Fits copula models to residuals and predictions of regression models.
    • 02_fit_copula_data.R: Fits copula models to income and wealth data and compare simulated data with original data.
  • /LM: Linear models fitting scripts.
    • 01_lm.R: Fits baseline LM models (main results).
    • 02_lm_cv.R: Fits LM models on different partitions of data and compute CV R-squared (robustness check).
    • 03_fit_lm_nonnested.R: Fit LM models that are not nested (robustness check).
  • /MARS: Scripts for fitting MARS models.
    • 01_mars.R: Fit MARS models (baseline as well as grouped by sign of income and wealth).
  • /RF: Random Forest model fitting and tuning scripts.
    • 01_rf_grid.R: Initial grid-search.
    • 02_rf_grid2.R: Refined grid-search using caret.
    • 03_rf_mbo.R: Hyperparameter tuning via model-based optimization (MBO).
    • 04_rf_fit_mbo.R: Fit final result of MBO and create plots.

Analysis of results

• /_analysis: Directory containing scripts for data analysis.
  • 01_combine_models.R: Script for combining models.
  • 02_sanity_checks.R: Sanity checks for the analysis.
  • 03_descriptive.R: First descriptive analysis of the data.
  • 04_analysis2.R: Analysis of the results. Depends on output of 01_combine_models.R.