Development of analysis methods using recent data a multivariate analysis of crash and naturalistic event data in relation to highway factors using the GIS framework.
Authors: Timothy J. Gordon, Lidia P. Kostyniuk, Paul E. Green, Michelle A. Barnes, Daniel F. Blower, Scott E. Bogard, Adam D. Blankespoor, David J. LeBlanc, B. R. Cannon, S. B. McLaughlin
This report documents and presents the results of a preliminary study into the use and validation of crash surrogates, to be obtained from naturalistic driving studies for the detailed analysis of risk factors. The approach is based on a unified statistical analysis of crash data and surrogate events using a spatial referencing system and a common measure of exposure. The specific crash type addressed in this study is single vehicle road departure crashes. It is proposed that suitable surrogates should be based on underlying continuous measures of disturbance in the driver's lateral control of the surrogate, while estimated time to road departure was found to show the correct statistical dependencies, consistent with the crash data. A number of additional enhancements to the crash surrogate definition were proposed, though larger data sets are neede to refine the analysis presented. Further exploratory analysis indicated that Extreme Value Theory is capable of giving plausible estimates for crash rates, provided a validated surrogate is used.