Finite mixture modeling approach for developing crash modification factors in highway safety analysis

Area of Science:

• Transportation Engineering
• Traffic Safety Analysis
• Statistical Modeling

Background:

• Crash Modification Factors (CMFs) are crucial for evaluating the safety impact of highway design and operational changes.
• Traditional models like the negative binomial (NB) may not fully capture the complexity of crash frequency data.
• The two-component finite mixture of negative binomial (FMNB-2) model offers a potential improvement for modeling crash data.

Purpose of the Study:

• To compare the performance of the NB model and the FMNB-2 model in developing CMFs.
• To derive and evaluate crash modification functions (CMFunctions) from both models.
• To assess the accuracy of combined CMFs and Adjustment Factors (AFs) derived from the FMNB-2 model.

Main Methods:

• Modeling crash data from rural multilane divided highways in California and Texas using both NB and FMNB-2 models.
• Deriving CMFunctions from the fitted models to represent the safety effects of covariates.
• Estimating combined CMFs for multiple treatments and developing AFs.

Main Results:

• The FMNB-2 model's CMFunction better reflects covariate effects and captures nonlinear relationships between covariates and safety.
• Combined CMFs derived from the FMNB-2 model are not simply multiplicative, indicating dependent safety effects.
• Adjustment Factors developed using the FMNB-2 model highlight potential over- or under-estimation by current Highway Safety Manual methods.

Conclusions:

• The FMNB-2 model provides a more robust framework for developing CMFs and AFs compared to the NB model.
• Safety analysts should consider utilizing FMNB-2 models for more accurate crash modification estimations.
• The findings suggest a need to re-evaluate current methods for combining CMFs, especially under complex covariate interactions.