An Automated Prior Robustness Analysis in Bayesian Model Comparison

Project for the third year class "Bayesian Statistics" at ENSAE IP Paris.

Benjamin Maurel
Gabriel Watkinson

Introduction

In this project, we implement the methods described in the paper An Automated Prior Robustness Analysis in Bayesian Model Comparison by Joshua Chan, Liana Jacobi and Dan Zhu, published in the Journal of Applied Econometrics in 2022.

A short report is associated with this repository, and can be found here, describing the methods and the results obtained.

Files and folders

The implementation is done in Python, and the code is available in the src folder.
The notebooks used to generate the results are available in the notebooks folder.

The data used

The data used in the project is available in the data folder. The dataset USdata_2019Q4.xlsx is taken from the Federal Reserve Bank of St. Louis' FRED-QD database and the sample period is 1948:Q1-2019:Q4. The first column of the file contains the dates. The second to last columns contain the values of GDP deflator, unemployment rate, real GDP growth and Fed funds rate, respectively.

The report and pdfs

The report is available in the pdfs folder, as well as the original paper and the instructions of the project.

For more ressources, please visit the author's website: http://joshuachan.org/research.html. You can directly download their Matlab code and the datasets used here : http://joshuachan.org/code/AD_ML_code.zip. The paper is also available : http://joshuachan.org/papers/AD_ML.pdf.

Running the code

If you want to rerun the experiments, you can clone the project and run the notebooks in the notebooks folder.

git clone https://github.com/gwatkinson/automated_prior_analysis.git

Reproducing the environment

The environment used to run the code is available in the poetry.lock file, it is generated with the pyproject.toml file.

If you have poetry installed, you can create the environment using the following command:

poetry install

This will create a virtual environment with all the dependencies taht are specified in the lock file, needed to run the code. It will also install the directory as a package, so that you can import the modules in the src folder.

Alternatively, you can install the dependencies manually using pip:

pip install -r requirements.txt

This will install the dependencies in your current environment.

Pre-commit

If you want to contribute, please install the pre-commit hooks (in the root folder with git and with the enviromnent activated):

pre-commit install

This installs hooks to /.git/hooks and run it once against the code:

pre-commit run --all-files

This will run some formatters and other hooks before each commit.

gwatkinson / automated_prior_analysis