考虑潜在类别的老年行人交通事故严重程度致因分析

人口老龄化问题日益突出，老年行人的出行安全同样引起各界重视。本文基于潜在类别聚类分析和随机参数Logit模型相结合的两步法深入探究影响老年行人交通事故严重程度的诱因。对北卡罗来纳州2007—2019年65岁及以上老年行人与机动车的碰撞数据进行清洗。为消除碰撞数据中固有的未观察到的异质性，首先进行潜在类别聚类分析，依据拟合优度指标确定最佳聚类数，将数据分成3个集群，分别对每个集群进行特征描述。然后，分别对每个集群建立随机参数Logit模型，以进一步探索集群内部未观察到的异质性，同时计算各显著变量的边际效应，量化其对事故严重程度概率的影响。结果显示，随机参数Logit模型具有更好的拟合优度；不同集群参数估计结果有所差异，一些变量只在特定集群内是显著的；集群1中，“救护车援助”为随机变量，集群2中“事故发生在城市”为随机变量，集群3中未发现随机变量，退化为多项Logit模型。本文研究结果可为交通工程师和政策制定者提供更可靠准确的老年行人交通事故严重程度诱因信息，为老年行人出行安全改善方案的制定提供理论支撑和技术支持。

Causes Analysis on Severity of Elderly Pedestrian Crashes Considering Latent Classes

The aging population is becoming more and more prominent, and the travel safety of elderly pedestrians is drawing attention. By proposing a two-step method that integrates latent class cluster analysis with random parameters Logit model, this paper explored the potential risk factors that contribute to the severity of elderly pedestrian crashes. We first cleaned the crash data of elderly pedestrians aged 65 and older with motor vehicles in North Carolina from 2007 to 2019. To eliminate the unobserved heterogeneity inherent in the crash data, a latent class cluster analysis was carried out to determine the optimal number of clusters based on the goodness of fit index. Three clusters of data were divided, and their characteristics were summarized. Then a random parameters Logit model was developed for each cluster to further explore the unobserved heterogeneity within the cluster, while the marginal effects of significant variables were calculated to quantify their impact on the probability of accident severity. The results show that the random parameters Logit model has better goodness of fit. The parameter estimates vary across clusters, and some variables are significant only within specific clusters. In cluster 1, "ambulance assistance" is a random variable, "accidents in the city" is a random variable in cluster 2, and no random variables are found in cluster 3, which degenerated to a multinomial Logit model. The results provide more reliable and accurate causes of traffic accidents in elderly pedestrians to traffic engineers and policymakers, supporting the formulation of safety plans for elderly pedestrians both in theory and technology.