Applying Bayesian hierarchical models to examine motorcycle crashes at signalized intersections

Haque, Md. Mazharul; Chin, Hoong Chor; Huang, Helai · 2009 · OpenAlex-citations

DOI: 10.1016/j.aap.2009.07.022

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the causal factors influencing motorcycle crashes at signalized intersections in Singapore, addressing the disproportionate vulnerability of motorcyclists in these environments. Motorcycles account for 36% of total crashes and 54% of fatalities in Singapore, with signalized intersections being particularly hazardous sites. The research aims to identify specific geometric and traffic characteristics that affect crash frequencies at both four-legged and T-intersections, filling a gap in literature that often aggregates all vehicle types or ignores intersection-specific dynamics for motorcyclists. The methodology employs Bayesian hierarchical models to analyze time-series cross-section panel data from 270 four-legged and 101 T-intersections over a four-year period (2003–2006). The dataset includes 1,948 crashes at four-legged intersections and 400 at T-intersections. The authors compared four statistical models: Poisson Gamma, Hierarchical Poisson Gamma, Hierarchical Poisson Lognormal, and Hierarchical Poisson Autoregressive lag-1 (AR-1). Model selection was conducted using the Deviance Information Criterion (DIC) and Predictive Loss Criteria (PLC), with parameter estimation performed via Markov Chain Monte Carlo (MCMC) methods. The analysis focused on intersection-level crash counts, incorporating variables such as lane count, median width, turn lane configurations, speed limits, traffic volume, and the presence of red-light cameras. The results indicate that the Hierarchical Poisson (AR-1) model provided the best fit for both intersection types, confirming significant temporal serial correlation in crash data. For four-legged intersections, the number of lanes on major and minor roadways significantly increased crash frequency, attributed to higher motorcycle exposure and accumulation at stop lines. The presence of a wide median (>2 meters) and uncontrolled left-turn lanes on major roadways also exacerbated crash risks by restricting driver visibility and creating complex conflict points. For T-intersections, exclusive right-turn lanes on both major and minor roadways, along with uncontrolled left-turn lanes on major roadways, were associated with increased crashes. High-speed roadways generally increased crash likelihood due to reduced reaction times. Conversely, the presence of red-light cameras significantly reduced motorcycle crashes for both intersection types by enforcing disciplined queuing and reducing early starts. The study concludes that specific geometric designs, particularly those increasing exposure or complexity, heighten motorcycle crash risks at signalized intersections. The findings highlight the effectiveness of red-light cameras as a countermeasure. By utilizing Bayesian hierarchical models, the research demonstrates the importance of accounting for site-specific effects and temporal correlations in safety analysis, providing a robust framework for identifying targeted interventions to improve motorcyclist safety.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• motorcycle crash typology
• cyclist safety
• vru crash typology
• helmet protective
• incidence prevalence

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes