Generalized nonlinear models for rear-end crash risk analysis.

Area of Science:

• Traffic Safety Engineering
• Transportation Modeling
• Statistical Analysis

Background:

• Traditional Generalized Linear Models (GLMs) for crash risk often assume linear relationships, which may not reflect real-world complexities.
• GLM limitations can lead to biased findings in highway safety research.
• Highway rear-end crashes remain a significant safety concern requiring advanced modeling techniques.

Purpose of the Study:

• To develop and apply a Generalized Nonlinear Model (GNM) for more accurate highway rear-end crash risk assessment.
• To investigate non-monotonic relationships between contributing factors and rear-end crash occurrence.
• To improve the understanding of factors influencing rear-end crash risk for better safety planning.

Main Methods:

• Utilized Washington State traffic safety data from 10 highway routes (2002-2006).
• Developed a GNM-based approach incorporating a nonlinear regression function.
• Compared GNM performance against traditional GLMs for rear-end crash risk modeling.

Main Results:

• The GNM approach effectively captured non-monotonic relationships, unlike GLMs.
• Identified parabolic impacts of factors like truck percentage and grade on crash risk.
• Demonstrated that GNM provides more nuanced and realistic explanations of crash contributing factors.

Conclusions:

• GNM offers superior flexibility and accuracy for modeling nonlinear relationships in crash data.
• Understanding parabolic impacts is crucial for effective rear-end crash safety improvement strategies.
• The GNM-based approach enhances the evaluation and selection of traffic safety interventions.