Bandwidth Selection for Kernel Density Estimation

Note: This repository only contains the code used for the estimation, simulation and visualisation parts of the thesis. The thesis itself is available upon request.

This project analyses the performance of different bandwidth selection methods when applied to both simulated and real world data. In particular, I apply the procedures to estimate the world income distribution which allows to investigate the development of income inequality between countries.

Building the Project

The project is written in Python and R. It is built using Waf. After a successful build, the full documentation of the project can be found in:

	project_documentation/index.html 

To run Waf and execute the files, you need to:

1. Save the project on your computer (clone the repository or save the zip file).

2. Install Miniconda or Anaconda in case they are not already installed and make sure that a LaTeX distribution can be found on your path.

3. Make sure an R executable is added to your path. Under Mac OS X, this can be achieved by opening the bash profile in a shell and adding for example:

    # R directory
    export PATH="${PATH}:/Applications/R.app/Contents/MacOS"
   

   Details on how to open the bash profile in a shell and general instructions for adding programmes permanently to your path for Windows, Mac and Linux can be found here.

4. Navigate to the project folder in a shell and execute the following commands to create a conda environment (named as the current directory) with a minimal Python setup.

   (Mac, Linux)

    source set-env.sh
   

   (Windows)

    set-env.bat
   

   Details for setting up a Python environment can be found here.

5. Execute the following commands in the shell:

    python waf.py configure
    python waf.py build
    python waf.py install
   

   The execution of the first command will fail if any of the programmes required to run the project is not installed.

Note

In case you just want to quickly execute the whole project, apply the following changes to greatly reduce the runtime:

src/model_specs/draws.json (line 2): "configuration_0": 10