Analysis on the heterogeneity of factors influencing the rear-end crash severity on highways

To explore the differences in the mechanisms leading to highway rear-end crash severity during daytime and nighttime, the correlated random parameter order Probit model with heterogeneity-in-means (CROPHM) was constructed to capture unobserved heterogeneity and explain the correlation between random parameters. A sample of 8 664 highway rear-end crashes was extracted from 2012-2018 Shanxi highway accidents dataset. With time variables, environmental variables and accident attributes as input indicators and the maximum injury severity as the output variable, the factors influencing the severity of rear-end crashes during the daytime and nighttime were explored separately. Firstly, likelihood ratio(LR)tests were conducted to validate whether rear-end crashes classified by daytime and nighttime should be modeled separately. Subsequently, two CROPHM models were established to examine the temporal differences in factors influencing the rear-end crash severity. Finally, through marginal effect analysis, corresponding recommendations were presented. The study findings indicate that rear-end crashes between 00:00 and 06:00 significantly increase the probability of fatal accidents by 0.063 7 compared with 18:00-24:00. Moreover, the probability of fatal crashes increases by 0.015 6 during nighttime multi-vehicle rear-end collisions compared with single-vehicle crashes. When large vehicles are involved in rear-end collisions, the probability of fatal accidents significantly increases, both during the daytime and nighttime. © 2024 Southeast University. All rights reserved.