基于随机参数Ordered Probit模型的山区公路摩托车事故严重程度研究  

Study on Severity of Motorcycle Accidents on Mountainous Highway Based on Random Parameter Ordered Probit Model

作　　者：李小刚 田毕江 杨文臣 李薇 岳松 LI Xiao-gang;TIAN Bi-jiang;YANG Wen-chen;LI Wei;YUE Song(Yunnan Transportation Investment Highway Construction First Engineering Co.,Ltd.,Kunming Yunnan 650034,China;Yunnan Transport Planning,Design and Research Institute Co.,Ltd.,Kunming Yunnan 650200,China)

机构地区：[1]云南交投公路建设第一工程有限公司,云南昆明650034 [2]云南省交通规划设计研究院有限公司陆地交通气象灾害防治技术国家工程实验室,云南昆明650200

出　　处：《公路交通科技》2023年第1期169-177,共9页Journal of Highway and Transportation Research and Development

基　　金：国家重点研发计划项目(2017YFC0803906);云南省交通运输厅科技项目(2019303);云南省基础研究计划项目(2019FB072);公司自立科技项目(ZL-2021-01)。

摘　　要：为准确识别山区双车道公路摩托车事故严重程度影响因素及其异质性效应,降低摩托车事故发生率及其严重度,综合考虑事故严重度有序性和影响因素异质性,以2012—2017年云南省楚雄州Y公路摩托车交通事故为研究对象,从驾驶员、车辆、道路、环境和事故特性五要素中选取16个潜在影响因素作为自变量,以摩托车事故严重程度作为因变量,分别构建了山区双车道公路摩托车事故严重程度固定参数有序probit模型(OP模型)和随机参数有序probit模型(RP-OP模型),并结合边际效应分析了各影响因素对摩托车事故严重程度的作用强度。结果表明:考虑个体异质性的RP-OP模型拟合优度高于OP模型,能准确刻画数据中未观测到的异质性和事故严重程度影响因素之间的交互作用,更适用于事故严重度影响机理的精细化研究;OP和RP-OP模型均识别出摩托车驾驶员性别、操作行为、是否有责、事故发生时间、光线条件、事故发生地点、路表情况、涉事车型显著影响摩托车事故严重程度的8个变量,其中女性摩托车驾驶员、驾驶员超车不当、涉及货车3个变量是分别服从(0.721,0.987^(2)),(0.432,0.344^(2)),(0.723,0.562^(2))正态分布的随机参数,对摩托车事故严重度具有显著的异质性影响效应。研究成果可为山区双车道公路摩托车事故预防提供决策参考,并针对性制定安全改善措施。In order to accurately identify the factors affecting the severity of motorcycle accidents and heterogeneity effects on two-lane mountainous highway,and reduce the accident rate and severity of motorcycle accidents,considering the order of accident severity and the heterogeneity of influencing factors,taking the motorcycle traffic accidents on the Y highway in Chuxiong Prefecture,Yunnan Province from 2012 to 2017 as the research object,selecting 16 potential influencing factors as the independent variables from the 5 elements(driver,vehicle,road,environment and accident characteristics),and the motorcycle accident severity as the dependent variable,the fixed parameter ordered probit model(OP model)and random parameter ordered probit model(RP-OP model)of motorcycle accident severity are constructed respectively,and the influence intensity of each influencing factor on the severity of motorcycle accidents is analyzed combining with marginal effect.The result shows that(1)The fitting goodness of the RP-OP model considering individual heterogeneity is higher than that of the OP model,it can accurately depict the interaction between unobserved heterogeneity in the data and the influencing factors of accident severity,which is more suitable for the refined study of the impact mechanism of accident severity.(2)Both the OP and RP-OP models identified 8 variables(gender,operation behavior,responsibility,accident time,light condition,accident location,road surface condition,and type of motorcycle driver)which significantly affecting the severity of motorcycle accidents),among them,3 variables(female motorcycle driver,improper overtaking,involving truck)are random parameters showed normal distribution,which have significant heterogeneous impact on motorcycle accidents.To be specific,the coefficient of female motorcycle driver’s obeys normal distribution(0.721,0.987^(2)),the coefficient of motorcycle driver’s improper overtaking obeys normal distribution(0.432,0.344^(2)),and the coefficients involving truck obey normal d

关 键 词：交通工程 摩托车事故严重程度 随机参数有序probit模型 山区双车道公路 异质性效应 

分 类 号：U491.31[交通运输工程—交通运输规划与管理]