Abstract
The Fully Bayesian (FB) approach to road safety analysis has been available for some time, but it is largely unevaluated and untested. This study is trying to bridge the gap by conducting a thorough evaluation of FB method for black spots identification and treatment effect analysis. First, an evaluation is conducted on the univariate FB versus the empirical Bayesian (EB) method for single level severity data through the development of various models, and multivariate FB versus univariate FB for multilevel severity data, as well as the performance of various ranking and evaluation criteria for black spots identification. It is confirmed that the FB method is superior to the EB with respect to key ranking criteria (expected rank, mode rank and median rank of posterior PM, etc.). The multivariate FB method is better than univariate FB for the multilevel severity crashes. Then a teat of the FB before-after method for treatment effect analysis is performed. Two FB testing frameworks were employed. First the univariate before-after fully Bayesian (FB) method was examined using three simulated datasets. Then multivariate Poisson log normal (MVPLN), univariate Poisson log normal (PLN) and PB (Poisson gamma) models were evaluated using two groups of California unsignalized intersections. Hypothetical treatment sites were selected from these datasets such that a significant effect would be estimated by the naive before-after method that does not account for regression to the mean. This study confirmed that FB methods can indeed provide valid results, in that they correctly estimate a treatment effect of zero at these hypothetical treatment sites after accounting for regression to the mean. Finally the EB and the validated FB before after methods were applied to evaluation of two treatments: the conversion of rural intersections from unsignalized to signalized control; and the conversion of road segments from a four-lane to a three-lane cross-section with two-way left turn lanes (also known as road diets). The result indicates that both FB and EB method can provide comparable treatment effect estimates. This would suggest it is still appropriate to conduct treatment effect analysis using the EB method for univariate crash data, but that it is essential in so doing to account for temporal trends in crash frequency.
Highlights
1.1 OVERVIEW OF ROAD SAFETY1.1.1 The Status of Road Safety Road traffic crashes constitute one of the world’s largest public health and injury problems
5.1 INTRODUCTION This chapter begins with a literature review of actual applications of the fully Bayesian (FB) method for hazardous site identification in road safety, and the FB ranking criteria that are used in these studies; this is followed by the the objectives, details and results of the evaluation study
The crash reduction rate (CRR) shows correctly that there are no significant treatment effects estimated by the FB method, suggesting that regression to the mean (RTM) has been properly accounted for, and that this method can be used for observational before-after studies
Summary
1.1 OVERVIEW OF ROAD SAFETY1.1.1 The Status of Road Safety Road traffic crashes constitute one of the world’s largest public health and injury problems. One of the most important tasks for road safety analysts, is explored in this Chapter This aspect of the research has recently been published (Lan et al, 2009; Lan and Persaud, 2010), and some of the documentation below is taken from those sources. We provide a detailed comparison and discussion of the pros and cons of the two Bayesian approaches (EB and FB), based on, and illustrated with, empirical applications These applications pertain to the evaluation of two treatments: the conversion of rural intersections from unsignalized to signalized control; and the conversion of road segments from a four-lane to a three-lane cross-section with two-way left turn lanes ( known as road diets). Part of The investigation of the conversion of rural intersections from unsignalized to signalized control has recently been published (Persaud et al, 2010; Lan et al, 2009) and some of the documentation below is taken from that source
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
