This repository contains the scripts used to reproduce the analyses in "A Bayesian workflow for securitizing casualty insurance risk" by Haines, Goold, & Shoun (2024).
The Bayesian models themselves are implemented in Stan scripts located in the bayesian-workflow-paper-2024/stan/ directory.
All analyses were run with Python 3.11. Once Python 3.11 is installed locally, we recommend the
following steps:
- navigate to your local
bayesian-workflow-paper-2024/directory - initialize a virtual environment:
python3.11 venv env - activate the environment:
source env/bin/activate - install requirements:
pip install -r requirements/requirements.txt
Analyses can then be reproduced by running the following scripts in order:
- download and pre-process the data:
python -m pull - run simulation-based calibration:
python -m sbc - run backtests along with prior and posterior predictive checks:
python -m backtest - run Balona (2020) working example:
python -m analyze
Once analyses are reproduced, figures are located in the bayesian-workflow-paper-2024/figures/ directory.