Evaluating the spatiotemporal clustering of traffic incidents

This research presents both theoretical results regarding the nature of spatiotemporal clustering on a network, and applied outcomes from examining such clustering with regard to traffic incidents. The analysis considers fatal traffic incidents in eastern Fairfax County, Virginia and injury incidents in Franklin County, Ohio. The spatiotemporal analytical methods of Knox and subsequent researchers are reviewed. Specific methods for performing spatiotemporal analysis are outlined, with special attention given to the interpretation of the results for traffic incidents. An argument is made for conducting spatial and temporal cluster analyses independently, in addition to spatiotemporal cluster analysis, a comparative analysis of methods for testing for the significance of spatiotemporal clusters is presented, and suggestions for delineating critical parameters for the Knox statistic are provided.

► Traffic incident analysis performed using temporal, spatial, and spatiotemporal clustering methods. ► Comparative analysis of significance testing methods including multiple-testing correction techniques. ► Methods for determining logical bounds on critical parameters for spatiotemporal cluster analysis.