Title

Exploring The Effect Of Different Neighboring Structures On Spatial Hierarchical Joint Crash Frequency Models

Abstract

Corridor safety analysis is a primary interest of many road safety studies. Corridors typically contain intersections and roadway segments. Having both components while analyzing corridors in addition to corridor-level variables in a hierarchical joint model would provide a comprehensive understanding of the existing corridor safety problems. There will probably be spatial correlation among road entities along a corridor, especially if the distance between the road entities is not large. Therefore, it is crucial to consider spatial effects in the model. However, this data structure is relatively new, and the best spatial weight matrix for this hierarchical spatial joint model has yet to be investigated. Therefore, this study estimates a hierarchical Poisson-lognormal (HPLN) joint model with spatial effects and explores the effect of different neighboring structures. A total of thirteen HPLN joint models are estimated: one HPLN joint model with corridor random effect and twelve HPLN joint models with spatial effects. Four types of conceptualization of spatial relationships were considered: (a) adjacency-based, (b) adjacency-route, (c) distance-order, and (d) distance-based spatial weight features. The results show the importance of incorporating spatial effects in the model. It was found that having a joint model is important since one of the best models is the adjacency-based first-order model, where the feeding road entities in addition to the directly adjacent road entity of the same type as the road entity of interest are considered. The results confirm the importance of spatial autocorrelation between road entities along the same corridor.

Publication Date

12-1-2018

Publication Title

Transportation Research Record

Volume

2672

Issue

38

Number of Pages

210-222

Document Type

Article

Personal Identifier

scopus

DOI Link

https://doi.org/10.1177/0361198118776759

Socpus ID

85048258282 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/85048258282

This document is currently not available here.

Share

COinS