This repository implements the econometric models presented in:
Wang, C., Abdel-Aty, M., & Han, L. (2025). Grouped random parameters Poisson-Lindley model with spatial effects addressing crashes at intersections: Insights from visual environment features and spatiotemporal instability. Analytic Methods in Accident Research, 100387. https://doi.org/10.1016/j.amar.2025.100387
These WinBUGS scripts fit the following count-data models (more in Codes):
- Grouped RPP-Lindley with spatial effects (primary specification in the AMAR paper)
- Grouped RPP-Lindley (no spatial effects)
- Grouped RPNB-Lindley (Poisson-Gamma as a reference “NB” specification)
While designed for crash counts at intersections, these routines can be applied to any nonnegative count outcome (e.g., hospital admissions, incident tallies).
- Macro- and micro-level spatial effects via conditional autoregressive (CAR) priors
- Random parameter (RP) grouping to capture unobserved heterogeneity
- Lindley mixing distribution for over-dispersion beyond Poisson
Note: The “NB” (negative binomial) implementations here use a Poisson–Gamma mixing representation to ensure direct comparability with Poisson-based specifications. See Geedipally et al. (2014) for guidance on DIC usage in crash models.
If you use these scripts or adapt them in any form, please cite:
Wang, C., Abdel-Aty, M., & Han, L. (2025). Grouped random parameters Poisson-Lindley model with spatial effects addressing crashes at intersections: Insights from visual environment features and spatiotemporal instability. Analytic Methods in Accident Research, 100387. https://doi.org/10.1016/j.amar.2025.100387
- Poisson–Gamma (NB) Lindley codebase by Amir Pooyan: https://github.com/Econometrics-in-r/Negative-Binomial-Lindley-Model-for-Excess-Zero-Data
- Model selection caution: Geedipally, S. R., Lord, D., & Dhavala, S. S. (2014). A caution about using deviance information criterion while modeling traffic crashes. Safety Science, 62, 495–498.