基于Scobit的交通事故受伤严重程度模型

Injury Severity Model of Motor Vehicle Crashes Based on Scobit

为了克服传统二项Logit(binary logit,BL)模型在累积概率等于0.5处斜率最大的默认假设,本研究构建了基于Scobit的交通事故受伤严重程度模型。该模型与BL模型相比,增加了偏斜系数,较传统BL模型更具通用性,有更广阔的应用范围。通过对交通事故受伤严重程度的建模估计,结果表明:Scobit模型与BL模型类似,且比BL模型在信息准则评价上表现更好,该模型可以成为交通事故受伤严重程度研究新的模型选择。

In order to overcome the default assumption that the slope is the greatest when the cumulative probability of the traditional Binary Logit(BL)model is equal to 0.5,an injury severity model of motor vehicle crashes based on Scobit model was established.Compared with the traditional BL model,the Scobit-based model was more versatile and with wider application,in which a skewness coefficient was added.By modeling and estimating the severity of traffic accident injuries,it is shown that Scobit model is similar to BL model,and better than BL model in the evaluation of information criteria.The proposed model can be a new model choice for the study of traffic accident injury severity.