Developing a grouped random parameters multivariate spatial model to explore zonal effects for segment and intersection crash modeling

Publication date: September 2018Source: Analytic Methods in Accident Research, Volume 19Author(s): Qing Cai, Mohamed Abdel-Aty, Jaeyoung Lee, Ling Wang, Xuesong WangAbstractIt is acknowledged that crash occurrence on segments and intersections could be affected by multilevel factors. Omission of important explanatory variables could result in biased and inconsistent parameter estimates. This paper contributes to the literature by examining the zonal effects which are always excluded or ignored in traffic safety research for segments and intersections. A grouped random parameters multivariate spatial model is proposed to identify both observable zonal effects and unobserved heterogeneity at the zonal level by considering the heterogeneous and spatial correlations. The proposed model is evaluated by comparing it with its three counterparts: a fixed parameters univariate spatial model without zonal factors, a random parameters univariate spatial model without zonal factors, and a random parameters univariate spatial model with zonal factors. The results indicate that the three random parameters models could consistently provide better performance than the fixed parameters model and the models including zonal factors outperform the models without zonal factors. Besides, the proposed model has the optimal model performance compared with its counterparts, which validates the concept of adopting the multivariate modeling framework to identify the heterogeneous and spatial correlatio...

Source: Analytic Methods in Accident Research - July 5, 2018 Category: Accident Prevention Source Type: research