M. Singh, W. Cheng, D. Samuelson, E. Clay, H. Tang
The aim of the study is to fill the research gap by developing the crash severity models for distinct crash types. The data are extracted from the Highway Safety Information System (HSIS) for the analysis of crashes of two severity levels (injury and non-injury) in the intersections of the state highway system in California. To identify significant variables of different crash types on the basis of crash severities, a mixed binary logit model was employed. The results demonstrate that the covariates have different effects on crash severities under various crash types. Among a large number of covariates, the number of vehicles involved, and rural roadway class demonstrate a positive significant impact on crash severities across five and four crash types, respectively. Following them, the urbanized intersections in the non-urban areas, the mainline number of lanes at the intersection and mainline signal mast arm exhibit modestly positive influence on crash severities under two crash types. On the contrary, some other variables showing a negative influence on crash severities were right or left independent alignment and cross-street number of lanes for three crash types, and lighted intersections for two crash types.
Keywords: injury severity; crash types; unobserved heterogeneity; category-wise severity model; mixed logit model