Bayesian Identification of High-Risk Intersections for Older Drivers via Gibbs Sampling

Abstract

Hierarchical Bayes methods are combined with an induced exposure model in order to identify intersections where the crash risk for a given driver subgroup is relatively higher than that for some other group. The necessary computations are carried out using Gibbs sampling, producing point and interval estimates of relative crash risk for the specified driver group at each site in a sample. The method is applied to data from 102 signalized intersections, and 10 were identified as showing high risk for older drivers. Left-turn crashes tended to predominate at these 10, whereas rear-end crashes were most common at geographically similar intersections not identified as showing high risk to older drivers.