Occupant injury severity in passenger car-truck collisions on interstate 80 in Wyoming: a Hamiltonian Monte Carlo Markov Chain Bayesian inference approach

Abstract

This paper investigates the significant contributing factors affecting the injury severity of car drivers, car passengers, and truck occupants from the aspects of car-truck collisions along interstate 80 (I-80) in Wyoming using a binary logit model with a Bayesian inference approach. Fixed- and random-effects models were developed to examine the effects on severe and non-severe injuries. Results showed that the random-effects model provided a better fit to the data than the fixed-effects model. Parameter estimates were sampled from the posterior distributions using No-U-Turn Hamiltonian Monte Carlo (NUT HMC) sampling method because of its efficiency over other Monte Carlo Markov Chain (MCMC) techniques. The analysis showed that occupant behavior characteristics, such as impaired driving, alcohol or illegal drugs, fatigue, and dangerous driving significantly increased the probability of injury severity. Unlit conditions, inclement weather, challenging roadway geometry, presence of junctions, downgrades, and curve sections were also likely to lead to more severe occupant injuries. The significant finding of this paper concludes that car drivers are more responsible than truck drivers contributing more severe injuries in car-truck collisions. Findings from this study are expected to help WYDOT and other related agencies take necessary actions and decide on management strategies in reducing car-truck collisions.