基于Ordered Probit模型的追尾事故严重性影响因素分析  

Analysis on Influencing Factors of Rear-end Crash Severity Based on Ordered Probit Model

作　　者：王鹏 卢小钊 闫章存[1] 张敏捷[2] WANG Peng;LU Xiao-zhao;YAN Zhang-cun;ZHANG Min-jie(School of Transportation ＆ Logistics , Southwest Jiaotong University , Chengdu Sichuan 610031;Ningbo University of Technology, Ningbo Zhejiang 315211, China)

机构地区：[1]西南交通大学交通运输与物流学院,四川成都610031 [2]宁波工程学院,浙江宁波315211

出　　处：《公路交通科技》2018年第4期102-107,122,共7页Journal of Highway and Transportation Research and Development

基　　金：浙江省教育厅科研项目(Y201533514)

摘　　要：为了减少机动车事故带来的人身财产损失,给相关管理部门提供决策意见的理论支持,分析追尾事故的独特规律,以及研究环境因素、道路因素、车辆因素和驾驶员因素等影响因素与追尾事故严重程度之间的致因关系至关重要。通过对北卡罗来纳州2010—2014年的交通事故数据进行筛选分析,最终得到1 315条具有完整信息的追尾事故数据,并对数据重新编码。考虑到事故严重性的分类具有次序的特性,研究拟采用Ordered Probit模型(ORP)对影响追尾事故严重程度的各种因素进行回归分析。为了找出具有统计学显著的影响因素,采用向后删除变量法筛选变量,从而得到了包含全部显著因素的模型。最终对模型进行了平行线检验和似然比检验,分析模型的统计学合理性,并选取指标进行了模型数据拟合度优劣的评价。研究结果表明:使用安全带、光照条件、道路线型条件、交通控制条件以及道路交通量这5个方面的因素与追尾事故严重程度的相关度较大。模型的边际影响值显示,正确使用安全带、良好的光照条件、合理的道路线型条件、适宜的交通控制措施能够显著降低追尾事故的严重性;而过大的交通量由于易产生多车追尾事故,事故严重性程度增加。显著性检验结果显示ORP模型对于分析追尾事故严重性及其影响因素具有明显作用。To reduce traveler casualties and property damages result from road crashes, and provide administrative department with theoretical support of decisions,it is critical to analyze the characteristic rule of rear-end crash and explore the causational relationships between environment factor, roadway factor,vehcle factor,driver factor and rear-end crash severity. Crash records from North Carolina in 2010-2014 are screened and analysed,and finally 1 315 complete rear-end crash records are remained and recoded. It is believed that the causation of crash severity is inherently ordinal,so an Ordered Probit（ORP） model is adopted for regression analysis of various factors affecting rear-end crash severities. A backward variable elimination method is adopted to find out the statistically significant factors,and a model containing all the significant factors is obtained. Finally,the linear hypotheses test and likelihood ratio test are conducted for the model to examine its statistic rationality,and a series of indices are selected to evaluate the fitness ofmodel data. The result shows that（1） the using of restraint,lighting condition,road alignment,traffic control condition and annual average daily traffic（AADT） is significantly related to rear-end crash severity.The model marginal effect value indicates that strict restraint usage, good light condition, rational road alignment,proper traffic control can significantly reduce the severity of rear-end crash;（2） because of the continuous collisions,heavy traffic can increase the crash severity. The significant test proves that ORP model is obviously instrumental for analyzing the rear-end crash severity and the influencing factors.

关 键 词：交通工程 道路安全 ORDERED PROBIT模型 追尾事故 严重程度 显著性检验 

分 类 号：U491.31[交通运输工程—交通运输规划与管理]