Related Experiment Videos
Applying the random effect negative binomial model to examine traffic accident occurrence at signalized
Hoong Chor Chin1, Mohammed Abdul Quddus
1Department of Civil Engineering, The National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore. cvechc@leonis.nus.edu.sg
Accident; Analysis and Prevention
|December 31, 2002
Summary
Traditional traffic accident models like Poisson and negative binomial (NB) have limitations. A random effect negative binomial (RENB) model better analyzes intersection safety by accounting for spatial and temporal factors.
Area of Science:
- Transportation Engineering
- Traffic Safety Research
- Statistical Modeling
Background:
- Poisson and negative binomial (NB) models are commonly used for traffic accident analysis at intersections.
- These models have limitations, including assumptions about variance-to-mean ratio and data independence, which are often violated due to unobserved heterogeneity and serial correlation.
- These violations make standard models inappropriate for accurately capturing complex accident patterns.
Purpose of the Study:
- To introduce and evaluate the random effect negative binomial (RENB) model as a superior alternative for analyzing traffic accident occurrence at intersections.
- To identify key factors influencing intersection safety using the RENB model.
- To assess the suitability of the RENB model through various goodness-of-fit statistics.
Main Methods:
- Utilized a time-series cross-section panel data approach to incorporate spatial and temporal effects.
- Applied the random effect negative binomial (RENB) model to analyze accident data from signalized intersections in Singapore.
- Employed goodness-of-fit statistics to validate the model's performance.
Main Results:
- The RENB model demonstrated suitability for analyzing intersection safety data.
- Eleven variables were found to significantly impact intersection safety.
- Key significant factors included total approach volumes, number of phases per cycle, uncontrolled left-turn lanes, and the presence of surveillance cameras.
Conclusions:
- The random effect negative binomial (RENB) model is a more appropriate statistical tool for analyzing traffic accident data at intersections compared to traditional Poisson and NB models.
- Understanding the influence of geometric, traffic, and control characteristics is crucial for enhancing intersection safety.
- Specific intersection features like traffic volume, signal phasing, and infrastructure elements (e.g., turn lanes, cameras) play a significant role in accident occurrence.