School of Civil Engineering and Transportation, South China University of Technology, Guangzhou 510641, China.
Jiangsu Province Collaborative Innovation Center of Modern Urban Traffic Technologies, Southeast University Road #2, Nanjing 211189, China.
Int J Environ Res Public Health. 2019 Jan 14;16(2):219. doi: 10.3390/ijerph16020219.
This study attempts to investigate spatial autocorrelation and spillover effects in micro traffic safety analysis. To achieve the objective, a Poisson-based count regression with consideration of these spatial effects is proposed for modeling crash frequency on freeway segments. In the proposed hybrid model, the spatial autocorrelation and the spillover effects are formulated as the conditional autoregressive (CAR) prior and the exogenous variables of adjacent segments, respectively. The proposed model is demonstrated and compared to the models with only one kind of spatial effect, using one-year crash data collected from Kaiyang Freeway, China. The results of Bayesian estimation conducted in WinBUGS show that significant spatial autocorrelation and spillover effects simultaneously exist in the freeway crash-frequency data. The lower value of deviance information criterion (DIC) and more significant exogenous variables for the hybrid model compared to the other alternatives, indicate the strength of accounting for both spatial autocorrelation and spillover effects on improving model fit and identifying crash contributing factors. Moreover, the model results highlight the importance of daily vehicle kilometers traveled, and horizontal and vertical alignments of targeted segments and adjacent segments on freeway crash occurrences.
本研究旨在探讨微观交通安全分析中的空间自相关和溢出效应。为了实现这一目标，提出了一种基于泊松分布的计数回归模型，考虑了这些空间效应，用于对高速公路路段的碰撞频率进行建模。在所提出的混合模型中，空间自相关和溢出效应分别被构造成条件自回归（CAR）先验和相邻路段的外生变量。利用中国开阳高速公路一年的碰撞数据，对所提出的模型进行了演示和与仅具有一种空间效应的模型进行了比较。贝叶斯估计在 WinBUGS 中进行的结果表明，高速公路碰撞频率数据中存在显著的空间自相关和溢出效应。与其他替代方案相比，混合模型的偏差信息准则（DIC）值较低，并且外生变量更为显著，这表明考虑空间自相关和溢出效应对于提高模型拟合度和识别碰撞影响因素具有重要意义。此外，模型结果强调了目标路段和相邻路段的每日交通量、水平和垂直对准对高速公路碰撞发生的重要性。
