RESUMEN
The identification of accident hot spots is a central task of road safety management. Bayesian count data models have emerged as the workhorse method for producing probabilistic rankings of hazardous sites in road networks. Typically, these methods assume simple linear link function specifications, which, however, limit the predictive power of a model. Furthermore, extensive specification searches are precluded by complex model structures arising from the need to account for unobserved heterogeneity and spatial correlations. Modern machine learning (ML) methods offer ways to automate the specification of the link function. However, these methods do not capture estimation uncertainty, and it is also difficult to incorporate spatial correlations. In light of these gaps in the literature, this paper proposes a new spatial negative binomial model which uses Bayesian additive regression trees to endogenously select the specification of the link function. Posterior inference in the proposed model is made feasible with the help of the Pólya-Gamma data augmentation technique. We test the performance of this new model on a crash count data set from a metropolitan highway network. The empirical results show that the proposed model performs at least as well as a baseline spatial count data model with random parameters in terms of goodness of fit and site ranking ability.
Asunto(s)
Accidentes de Tránsito/prevención & control , Entorno Construido/clasificación , Administración de la Seguridad/métodos , Teorema de Bayes , Humanos , Modelos Estadísticos , Seguridad , Análisis EspacialRESUMEN
Permeable friction course (PFC), a porous hot-mix asphalt, is typically applied to improve wet weather safety on high-speed roadways in Texas. In order to warrant expensive PFC construction, a statistical evaluation of its safety benefits is essential. Generally, the literature on the effectiveness of porous mixes in reducing wet-weather crashes is limited and often inconclusive. In this study, the safety effectiveness of PFC was evaluated using a fully Bayesian before-after safety analysis. First, two groups of road segments overlaid with PFC and non-PFC material were identified across Texas; the non-PFC or reference road segments selected were similar to their PFC counterparts in terms of site specific features. Second, a negative binomial data generating process was assumed to model the underlying distribution of crash counts of PFC and reference road segments to perform Bayesian inference on the safety effectiveness. A data-augmentation based computationally efficient algorithm was employed for a fully Bayesian estimation. The statistical analysis shows that PFC is not effective in reducing wet weather crashes. It should be noted that the findings of this study are in agreement with the existing literature, although these studies were not based on a fully Bayesian statistical analysis. Our study suggests that the safety effectiveness of PFC road surfaces, or any other safety infrastructure, largely relies on its interrelationship with the road user. The results suggest that the safety infrastructure must be properly used to reap the benefits of the substantial investments.
Asunto(s)
Accidentes de Tránsito/prevención & control , Conducción de Automóvil , Planificación Ambiental , Seguridad , Tiempo (Meteorología) , Accidentes de Tránsito/estadística & datos numéricos , Teorema de Bayes , Fricción , Humanos , Modelos Estadísticos , Porosidad , Análisis de Regresión , Propiedades de Superficie , TexasRESUMEN
Crash statistics suggest that horizontal curves are the most vulnerable sites for crash occurrence. These crashes are often severe and many involve at least some level of injury due to the nature of the collisions. Ensuring the desired pavement surface condition is one potentially effective strategy to reduce the occurrence of severe accidents on horizontal curves. This study sought to develop crash injury severity models by integrating crash and pavement surface condition databases. It focuses on developing a causal relationship between pavement condition indices and severity level of crashes occurring on two-lane horizontal curves in Texas. In addition, it examines the suitability of the existing Skid Index for safety maintenance of two-lane curves. Significant correlation is evident between pavement condition and crash injury severity on two-lane undivided horizontal curves in Texas. Probability of a crash becoming fatal is appreciably sensitive to certain pavement indices. Data suggested that road facilities providing a smoother and more comfortable ride are vulnerable to severe crashes on horizontal curves. In addition, the study found that longitudinal skid measurement barely correlates with injury severity of crashes occurring on curved portions. The study recommends exploring the option of incorporating lateral friction measurement into Pavement Management System (PMS) databases specifically at curved road segments.