Spatial joint analysis for zonal daytime and nighttime crash frequencies using a Bayesian bivariate conditional autoregressive model

This study presents a joint analysis of daytime and nighttime crash frequencies at the zone level with consideration of spatial correlations. Crash data from 131 traffic analysis zones in Hong Kong in 2011 are investigated. A Bayesian bivariate conditional autoregressive model is proposed to establi...

Full description

Saved in:
Bibliographic Details
Published in:Journal of transportation safety & security 2020-04, Vol.12 (4), p.566-585
Main Authors: Zeng, Qiang, Wen, Huiying, Wong, S.C., Huang, Helai, Guo, Qiang, Pei, Xin
Format: Article
Language:English
Subjects:
Autoregressive models
Bayesian analysis
Bayesian inference
Bivariate analysis
bivariate conditional autoregressive model
Crashes
Daytime
daytime and nighttime
Dependent variables
Land use
Night
Nighttime
Parameter estimation
Residential areas
Roads
Spatial analysis
spatial correlation
Traffic
Zonal safety
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study presents a joint analysis of daytime and nighttime crash frequencies at the zone level with consideration of spatial correlations. Crash data from 131 traffic analysis zones in Hong Kong in 2011 are investigated. A Bayesian bivariate conditional autoregressive model is proposed to establish links between crash frequencies and traffic attributes, road network characteristics, and land use patterns. The proposed model allows not only for the distinct heterogeneous and spatial effects of each dependent variable, but also for the correlations between them. The parameter estimates indicate that more daytime and nighttime crashes are associated with more vehicle hours traveled and with networks that have greater global integration. Average speed alone has a significant negative effect on daytime crashes. The crash risk in commercial and other areas is lower than that in residential areas, but the crash risk in areas of mixed residential and commercial use is higher. Meanwhile, significant spatial autocorrelation emerges across zones and explains 46.7% and 48.2% extra-Poisson variations for daytime and nighttime crash frequencies, respectively. High positive correlations are found in both heterogeneous and spatial effects. These findings, together with its better performance on model fit than the univariate counterparts, demonstrate the strength of the proposed model.
ISSN:1943-9962
1943-9970
DOI:10.1080/19439962.2018.1516259