Analysis of Road Crash Frequency with Spatial Models

Abstract

Despite the evident spatial character of road crashes, limited research has been conducted in road safety analysis to account for spatial correlation; further, the practical consequences of this omission are largely unknown. The purpose of this research is to explore the effect of spatial correlation in models of road crash frequency at the segment level. Different segment neighboring structures are tested to establish the most appropriate one in the context of modeling crash frequency in road networks. A full Bayes hierarchical approach is used with conditional autoregressive effects for the spatial correlation terms. Analysis of crash, traffic, and roadway inventory data from a rural county in Pennsylvania indicates the importance of including spatial correlation in road crash models. The models with spatial correlation show significantly better fit to the data than the Poisson lognormal model with only heterogeneity. Parameters significantly different from zero included annual average daily traffic (AADT) and shoulder widths less than 4 ft and between 6 and 10 ft. In four models with spatial correlation, goodness of fit was improved compared with the model including only heterogeneity. More important yet is the potential of spatial correlation to reduce the bias associated with model misspecification, as shown by the change in the estimate of the AADT coefficient and other parameters.

Impact of traffic congestion on road accidents: A spatial analysis of the M25 motorway in England

“…The elasticity of AADT appears a little high in this study compared with some of the previous studies which reported that the elasticity ranges from 0.6 -0.7 (e.g. Abdel-Aty and Radwan, 2000; Bird and Hashim, 2006;Aguero-Valverde and Jovanis, 2008). This may be because these studies were undertaken under different road conditions.…”

“…The Poisson based models include the Poisson-lognormal model (e.g. Lord and Miranda-Moreno, 2007;Aguero-Valverde and Jovanis, 2008); and the Poisson-gamma model (also known as Negative Binomial model) which is widely used in accident analysis (e.g. Shankar et al, 1995;Milton and Mannering, 1998;Abdel-Aty and Radwan, 2000;Lord, 2000;Amoros et al, 2003;Noland and Quddus, 2004;Kim et al, 2006;Haynes et al, 2007).…”

“…This method was initially used in ecological analysis and disease mapping (e.g. Ghosh et al, 1999;MacNab and Dean, 2001), and then recently in road accident analysis (Miaou et al, 2003;Aguero-Valverde and Jovanis, 2006;Li et al, 2007;Aguero-Valverde and Jovanis 2008). According to Jovanis (2006 and2008), compared with non-spatial Poisson models, spatial models can generally produce consistent results and in some cases can better fit the data.…”

“…Poisson based models, such as a Poisson-lognormal, a Poisson-gamma and a Poisson-lognormal with conditional autoregressive (CAR) priors have been recommended and widely used in the literature (e.g. Shankar et al, 1995;Milton andMannering, 1998, Abdel-Aty andRadwan, 2000;Lord, 2000;Amoros et al, 2003;Miaou et al, 2003;Kim et al, 2006;Lord and Miranda-Moreno, 2007;Aguero-Valverde and Jovanis 2008;Quddus, 2008). The base form of Poisson based model can be expressed as follows:…”

“…Bird and Hashim (2006) used a sample of rural two-lane single carriageways. A similarly study undertaken by Aguero-Valverde and Jovanis (2008) was also based on rural two-lane roads. Mitra et al (2007) further showed that the effects of AADT on different roads are different while looking at junction accidents, and it was found that generally the coefficient of AADF for "major" roads is higher than "minor" roads.…”