Title

Transferring And Calibrating Safety Performance Functions Among Multiple States

Keywords

Calibration; Highway safety manual; Negative binomial regression; Safety performance functions; Transferability

Abstract

Safety performance functions (SPFs), statistical regression models, by predicting traffic crash counts by crash type, severity and facility type, aid traffic engineers in the process of identifying high frequency crash locations. Developing SPFs requires the collection and processing of traffic, crash, road design and other characteristics’ data. Jurisdictional agencies may choose not to develop their own SPFs and cut back on their resources by adopting SPFs provided by the national Highway Safety Manual (HSM). The HSM also provides a technique to calibrate its SPFs to the specific jurisdictions’ conditions. Yet, the technique is subject to criticism. This research is aimed at exploring the transferability of SPFs of rural divided multilane highway segments of Florida, Ohio, Illinois, Minnesota, California, Washington and North Carolina. The SPFs are negative binomial (NB) models as are those provided by the HSM. We address the fault of instinctively applying the HSM's SPFs to a particular locality without verifying whether the SPFs are transferable to the locality and compare different states’ crash patterns. Remarkably, it is found that specific SPFs of Ohio, Illinois, Minnesota and California are transferable to either of the four states. In addition, in this research, a calibration technique is proposed as an alternative to that of the HSM and two other calibration methods introduced in the traffic safety literature. They are the calibration function and the calibration of the transferred model's constant term in conjunction with the over-dispersion parameter. Our proposed calibration technique, namely local regression, is demonstrated to be more reliable than the HSM's and the ones previously proposed in the literature.

Publication Date

8-1-2018

Publication Title

Accident Analysis and Prevention

Volume

117

Number of Pages

276-287

Document Type

Article

Personal Identifier

scopus

DOI Link

https://doi.org/10.1016/j.aap.2018.04.024

Socpus ID

85046626914 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/85046626914

This document is currently not available here.

Share

COinS