英语论文网

orida, 64.2% of left-turn crashes involved injury,
∗ Corresponding author at: Department of Civil & Environmental Engineering,
University of Central Florida, Engr II, Room 301B, Orlando, FL 32816-2450,
United States. Tel.: +1 407 8234902; fax: +1 407 8234676.
E-mail address: xuewang@mail.ucf.edu (X.Wang).
whereas the percentage of injury crashes was only 50.1% for allother crashes.
From 2002, a series of crash frequency studies have beenconduced in Florida to identify the crash profiles for themajor intersectiontypes (Abdel-Aty and Wang, 2006; Wang and Abdel-Aty,
2006, 2007, 2008;Wang et al., 2006). In one study,Wang and Abdel-Aty (2008) investigated conflicting flows, intersection geometricdesign features, and traffic control and operational features on leftturn
crash occurrence. Left-turn crasheswere classifiedinto distinctconflicting patterns (i.e., left-turn traffic colliding with opposingthrough traffic, or with near-side through traffic, etc.), and then thecrash frequencies of different patterns were modeled. The studiesindicate there are obvious differences in the factors which correlatedwith different left-turn collisions. However, crash frequencystudies model accumulated crash counts, which ignores the differenceof severe and minor crashes. Therefore, they are unable toinvestigate how specific features affect crash injury severity.
The left-turn crashes at signalized intersections result in a hugecost to society in terms of death, injury, lost productivity, and propertydamage. However, how the different factors affect left-turn
0001-4575/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aap.2008.06.001X. Wang, M. Abdel-Aty / Accident Analysis and Prevention 40 (2008) 1674–1682 1675crash severity is still not clear. For example, traffic volume hasbeen identified as the most significant factor affecting crash occurrence
(Wang and Abdel-Aty, 2008), but it is not clear whether trafficflow affects crash severity. Left-turn phase has been identified to
be significant for left-turn crash occurrence, but no study investigatesits influence on crash severity. The purpose of this study is toinvestigate how traffic characteristics, driver attributes, vehicular
characteristics, roadway geometry features, environmental factors,and crash characteristics affect left-turn crash injury severity.In police reports, crash injury is categorized into five levelsbased on the most serious injury to any person involved in a crash:no injury, possible injury, non-incapacitating injury, incapacitatinginjury and fatal injury. Multinomial logit models were specifiedfor multiple alternatives of severity. Shankar and Mannering(1996) considered environmental, roadway, vehicular, and rider#p#分页标题#e#
characteristics in their multinomial logit analysis of motorcycleriderseverity on single-vehicle motorcycle crashes. Carson andMannering (2001) developedmultinomial logit models to examinethe effect of ice-warning signs on crash severity for different roadwayfunctional classes. Ulfarsson and Mannering (2004) exploreddifferences in severity between male and female drivers in singleand two-vehicle collisions; separate multinomial logit models ofseverity were estimated for male and female drivers. However, thelogit model’s assumption of independent errors for each alternative
is inconsistent with the fact that the alternatives for crash injuriesare ordered.With ordered alternatives