This presentation reports on a recent full Bayes analysis of the safety effectiveness of four-lane to three-lane conversions. The study was performed as part of a creative component by a graduate student, Wen Li, working with Dr. Alicia Carriquiry of the Iowa State University Department of Statistics and Dr. Michael Pawlovich of the Iowa Department Transportation Office of Traffic and Safety. The Iowa study refutes a recent Federal Highway Administration (FHWA) Highway Safety Information System (HSIS) study that found little safety benefit from four-lane to three-lane conversions. The FHWA study fitted negative binomial regression models to the crash frequencies at each project site. The Iowa study was based on a much richer database and a superior statistical analysis. This was the first time a full Bayes analysis has been applied to a safety countermeasure. The full Bayes analysis shows some very distinct advantages over the results obtained from fitting negative binomial regression models.
Abstract