Citation

Lu J, Haleem K, Alluri P, Gan A. Transp. Res. Rec. 2013; 2398: 83-92.

Copyright

(Copyright © 2013, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.3141/2398-10

PMID

unavailable

Abstract

The empirical Bayes approach adopted in the Highway Safety Manual and the SafetyAnalyst software application require the use of safety performance functions (SPFs). SafetyAnalyst adopts a form of SPF, known as the simple SPF, that relates crash experience to traffic volume only. It is a flow-only model that is calibrated by using all sites irrespective of their base geometric conditions. Full SPFs, however, relate crash occurrence to roadway geometric characteristics in addition to traffic characteristics. This study compared the simple SPFs provided in SafetyAnalyst with full SPFs in two safety applications: crash prediction performance and identification of high-crash locations. To compare the prediction performance, the simple and full SPFs were estimated with data collected on urban four-lane freeway interchange influence areas in Florida. Models were estimated for both total crashes and fatal and injury crashes. The mean absolute deviance and the mean square prediction error were used to assess and compare the prediction performance of the two models, and the variation in ranking the high-crash locations with each model was also examined. The results showed that the two models yielded very similar performance of crash prediction and network screening. This empirical result supports the use of the flow-only SPF model adopted in SafetyAnalyst, whose development requires much less effort compared with that for full SPFs.