The focus of this project was to identify and prepare crash, near-crash, and baseline data sets extracted from the Second Strategic Highway Research Program (SHRP2) Naturalistic Driving Study (NDS) trip files, then to make that information available to researchers for use in their analysis projects. A dozen trigger algorithms were executed on 5,512,900 trip files in the SHRP2 NDS, and a manual validation of these algorithms identified 1,549 crashes and 2,705 near-crashes. A longitudinal deceleration-based algorithm produced the highest percentage of valid crashes and near-crashes. Baselines were selected via a random sample stratified by participant and proportion of time driven. Triggered epochs and the resulting crashes and near-crashes were reviewed and analyzed by a large team of data reductionists and quality control coordinators following a rigorous training, testing, and monitoring protocol. As a result, 20,000 baselines, including all drivers in the SHRP2 NDS, were prepared and are recommended for researchers using a case-cohort design. An additional 12,586 baselines are also available for researchers who may require more power in their analyses but are able to forego a fully proportional representation of all drivers in the study. Researchers using this data set are encouraged to review the data dictionaries on the InSight website prior to doing analysis and to be particularly careful in selecting the best subset of crashes, near-crashes, and baselines that informs their research questions. (Author/publisher)
Abstract