F-Divergence Robust Risk Toolkit
This project implements a robust risk assessment toolkit in R based on F-divergences. It reproduces and extends ideas from the literature and applies them in both simulation-based and practical examples.
Highlights
- Implemented robust risk calculations under divergence-based uncertainty sets.
- Worked with lognormal and Weibull reference models.
- Studied applications involving inventory pooling, dependence modelling, and Hong Kong COVID-19 case counts.
Methods
- Built functions to estimate worst-case expectations under divergence constraints.
- Reproduced core experiments from the reference framework and extended them to additional settings.
- Used simulation and comparative analysis to evaluate robustness under model uncertainty.
Findings
- The project illustrates how divergence-based methods can provide a practical way to stress-test risk measures.
- It also shows how robust procedures can be adapted to applied settings beyond purely theoretical examples.