This repository contains code, data, and results from a simulation study on the properness of scoring rules in survival analysis, including supplementary experiments and visualizations.
See HTML report.
- Weibull distribution variability [link] (main manuscript; Figure 1)
- Empirical violations of properness [link] (main manuscript; Table 1)
- SBS improperness: Bias and Tail Effects [link] (main manuscript; Figures 2,3)
- SBS improperness: Administrative Censoring [link] (main manuscript; investigation related to the experiments section)
- Empirical violations of properness using G(t) [link] (main manuscript; Table C1)
- Real-World Benchmark Evaluation [link] An exploratory comparison of survival metrics including 3 scoring rules (RCLL, RCLL*, ISBS), the C-index and D-calibration, on a few real-world survival datasets.
Sonabend, R., Zobolas, J., De Bin, R., Kopper, P., Burk, L., & Bender, A. (2024). Examining properness in the external validation of survival models with squared and logarithmic losses. https://arxiv.org/abs/2212.05260v3