Two vehicle truck-involved crash severity model using ordered logit regression

Abstract

Traffic crash is always a big issue for traffic safety researchers. The crash study is often focused on crash frequency and severity. This study considered the 2 vehicle truck involved crash happened in CA Bay area Interstate 880 from year 2005 to 2009 and divided to dataset into two parts: non-at-fault truck driver crash and at-fault truck driver crash. The crash information sources include SWITRS, PeMS and TASAS geometry. The ordered logit regression method is used to model the two data sets. After modeling we find that these 15 variables Day Of Week, Crash type, Safety measures, Road surface, Volume, V-Sd, S-Sd, Occupancy, Median width, Sex, Cellphone usage, Movement, Weather, Lighting, Surface type all affect the crash severity, while compared with the result for the at-fault truck drivers, only these 7 variables: Crash type, Sobriety, Shoulder width, Lighting, Weather, Speed, S-Sd affect crash severity in the same significance level. For non at fault truck drivers crash, Being female, Cellphone hands free in use, Weekends (Saturday and Sunday), Proceeding straight, Sideswipe crash type, Wet road surface, Dusk-Dawn, Concrete road surface type all tend to increase the crash severity level. For at fault truck drivers, Higher Standard deviations of speed (S-SD), Higher Speed, Broadside crash, Dusk-Dawn, Driving after drinking are found to increase the crash severity level.