Investigation of driver injury severities in rural single-vehicle crashes under rain conditions using mixed logit and latent class models

被引:82
|
作者
Li, Zhenning [1 ]
Ci, Yusheng [2 ]
Chen, Cong [3 ]
Zhang, Guohui [1 ]
Wu, Qiong [1 ]
Qian, Zhen [4 ]
Prevedouros, Panos D. [1 ]
Ma, David T. [1 ]
机构
[1] Univ Hawaii Manoa, Dept Civil & Environm Engn, 2500 Campus Rd, Honolulu, HI 96822 USA
[2] Harbin Inst Technol, Dept Transportat Sci & Engn, 73 Huanghe Rd, Harbin 150090, Heilongjiang, Peoples R China
[3] Univ S Florida, Ctr Urban Transportat Res, 4202 East Fowler Ave,CUT100, Tampa, FL 33620 USA
[4] Carnegie Mellon Univ, Civil & Environm Engn, Pittsburgh, PA 15213 USA
来源
关键词
Driver injury severity; Mixed logit model; Latent class model; Unobserved heterogeneity; Temporal instability; RANDOM PARAMETERS APPROACH; MULTINOMIAL LOGIT; CLUSTER-ANALYSIS; ORDERED PROBIT; HETEROGENEITY; AGE; DETERMINANTS; CONSUMPTION; PATTERNS; WEATHER;
D O I
10.1016/j.aap.2018.12.020
中图分类号
TB18 [人体工程学];
学科分类号
1201 ;
摘要
Due to limited visibility and low skid resistance on road surface, single-vehicle crashes under rain conditions, especially those occurred in rural areas, are more likely to result in driver incapacitating injuries and fatalities. A three-year crash dataset including all rural single-vehicle crashes under rain conditions from 2012 to 2014 in four South Central states, i.e., Texas, Arkansas, Oklahoma, and Louisiana, are selected in this paper to analyze the impact factors on driver injury severity. The mixed logit model (MLM) and the latent class model (LCM) are developed on the same dataset. Several parsimony indices, e.g., AIC and BIC, and as well as McFadden pseudo r-squared, are calculated for all the models to evaluate their respective performance. Results show that choosing the uniform distribution as the prior for random parameters could better improve the goodness-of-fit of the MLM than using normal and lognormal distributions. In addition, the two-class LCM also shows superiority when compared to three- and four-class LCMs. Finally, a careful comparison between these two models is conducted, and the results indicate that the LCM has a slightly better performance in analyzing the aforementioned dataset in this study. Model estimation results show that curve, on grade, signal control, multiple lanes, pickup, straight, drug/alcohol impaired, and seat belt not used can significantly increase the probability of incapacitating injuries and fatalities for drivers in the two models. On the other hand, wet, male, semi-trailer, and young can significantly decrease the probability of incapacitating injuries and fatalities for drivers. This study provides an insightful understanding of the effects of these attributes on rural single-vehicle crashes under rain conditions and beneficial references for developing effective countermeasures for severe injury prevention.
引用
收藏
页码:219 / 229
页数:11
相关论文
共 31 条
  • [31] A Motorcyclist-Injury Severity Analysis: A Comparison of Single-, Two-, and Multi-Vehicle Crashes Using Latent Class Ordered Probit Model
    Li, Jing
    Fang, Shouen
    Guo, Jingqiu
    Fu, Ting
    Qiu, Min
    ACCIDENT ANALYSIS AND PREVENTION, 2021, 151