A binomial distribution model for describing pedestrian-vehicle crashes in urban areas

Abstract

This study aims to investigate the relationship between the numbers of pedestrian-vehicle crashes (PVCs) and the traffic and built environments from a macroscopic perspective. A binomial distribution model has been developed to represent the occurrence of PVCs. To calibrate the model, Bayesian analysis using the Markov chain Monte Carlo method has been employed. The results identify thirteen variables representing urban activities and traffic conditions, including land uses, degrees of mixed use, points of interest, various passage widths, and street hierarchy that significantly impact PVCs. Additionally, the study unveils the spatial distribution of PVC probabilities and exposures. This research contributes to the field by developing an analytical framework for comprehending PVCs from a macroscopic viewpoint, introducing innovative methods for uncovering latent variables, integrating diverse types of data into the analysis, and creating a model for simulating the effects of urban planning revisions and traffic management strategies.