A nested grouped random parameter negative binomial model for modeling segment-level crash counts

Area of Science:

• Transportation Engineering
• Statistical Modeling
• Road Safety Analysis

Background:

• Road safety analysis requires accurate crash prediction models.
• Existing models may not fully capture complex correlations in crash data over time and across geographical units.
• Longitudinal data analysis is crucial for understanding temporal trends in traffic accidents.

Purpose of the Study:

• To propose and validate a novel nested grouped random parameter negative binomial framework for modeling crash counts.
• To account for correlations in crash data along county routes and over time.
• To analyze crash count trends on undivided two-lane arterial roads in Ohio from 2012 to 2017.

Main Methods:

• Development of a three-level longitudinal framework incorporating random parameters.
• Application of the model to crash data from Ohio's undivided two-lane arterial roads.
• Comparison of model variants with fixed and varying slopes to assess goodness-of-fit.

Main Results:

• A significant quadratic relationship between time and crash count was observed, indicating a decreasing rate of increase.
• 17% of segments and 2% of routes showed crash counts decreasing at an accelerating rate over time.
• Segment length positively correlated with crashes, while total shoulder width showed a negative correlation, with significant variations across routes.

Conclusions:

• The proposed nested grouped random parameter negative binomial model offers high forecast accuracy for crash counts.
• The model provides valuable insights into temporal trends and spatial variations in road safety.
• This framework serves as a robust tool for data-driven decision-making in road safety improvement strategies.