Bayesian approaches on observational before-after highway safety analysis

Observational before-after evaluation of roadway safety treatments effect is a necessary and challenging study area for traffic safety engineers. In decades, empirical Bayesian approach has been widely adopted among researchers in this area. In this approach, the researcher combined before period observation data with a safety performance function estimated from similar groups to estimate the expected number of crashes that would have occurred in the after period without the treatment. Comparing this estimate with the number of crashes observed in the after period, the researcher evaluated the effectiveness of the treatment. This approach accounts for well-known regression-to-the-mean effects. However, as a result of the recent progress in numerical methods made in recent years, more attention has been given on the fully Bayesian approach. The fully Bayesian approach implemented via hierarchical models is believed to have more advantages over the empirical Bayesian approach. The fully Bayesian approach required less data for untreated reference sites, it better accounts for uncertainty in data used, and it provides more detailed causal inferences and more flexibility in selecting crash count distributions. The objective of this study is to compare the differences of the empirical Bayesian procedure and the fully Bayesian procedure for before-after road safety evaluation studies. An illustrative example is included with a step-by-step procedure for both approaches. Throughout three approaches in the example, the estimated safety effects are calculated for MODOT\u27s reference --Abstract, page iii