Analysis of crash rates and surrogate events Unified Approach.
Authors: Tim J. Gordon, Lidia P. Kostyniuk, Paul E. Green, Michelle A. Barnes, Daniel Fredrick Blower, Adam D. Blankespoor, Scott E. Bogard.
A preliminary study was done into the use and validation of crash surrogates, which are 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 that uses a spatial referencing system and a common measure of exposure. Statistical methods based on a bivariate response and Bayesian update models were adapted to the joint analysis of crashes and surrogates. The study specifically addresses road-departure crashes involving a single vehicle. It is proposed that suitable surrogates be based on underlying continuous measures of disturbance in the driver-s lateral control of the vehicle. Naturalistic driving data from a field operational test conducted in southeastern Michigan were spatially joined with highway data and crash data from the same area, and a set of candidate crash surrogates was tested. Analysis results indicated that simple lateral lane position did not provide a satisfactory surrogate, whereas estimated time to road departure was found to show the correct statistical dependencies, consistent with the crash data. The approach developed in the study provided a way to assess crash risk in a common framework and also to validate or invalidate candidate surrogates. When applied to data from the future SHRP 2 naturalistic driving study, the increased statistical power resulting from the much larger data set will provide more definitive conclusions about surrogate validity and factors influencing overall crash risk.