Citation

Zeng Q, Wang F, Wang Q, Pei X, Yuan Q. Transp. Lett. 2023; 15(6): 553-560.

Copyright

(Copyright © 2023, Maney Publishing, Publisher Informa - Taylor and Francis Group)

DOI

10.1080/19427867.2022.2072459

PMID

unavailable

Abstract

This paper develops a multivariate spatial model for the joint analysis of daytime and nighttime crash frequencies by injury severity in traffic analysis zones (TAZs). The model specification allows for spatial correlation across TAZs, heterogeneous effects specific to crash time and severity, and correlations across response types under a Bayesian multivariate conditional autoregressive framework. One hundred and thirty one TAZs in Hong Kong, China, with traffic crash, traffic flow, roadway network, and land use data for a one-year period are selected to calibrate the advocated model. Considerable spatial and heterogeneous effects are found for each type crash frequency. Significant correlations exist in the heterogeneous effects for various severity levels and those for daytime and nighttime. The Bayesian estimates of the regression coefficients reveal that there are significant inconsistencies in the set of factors contributing to zonal daytime and nighttime crashes at various severity levels.

Language: en

Keywords

bayesian inference; conditional autoregressive prior; multivariate spatial model; spatial correlation; Transportation safety planning; zonal transportation safety