Factors Impacting the Likelihood of Near-Crashes and Crashes

Project Description This study will build a nested logit model for predicting the likelihood near-crashes and crashes by incident type. The upper-level nest will comprise three alternatives (a) no-risky event, (b) near-crash, and (c) crash. The NDS dataset classifies the trips into finer categories but we choose this aggregate three-way classification for our study. Based on preliminary analysis of InSight data, there are over 28,000 trips (including only the "balanced sample baseline trips") with about 1800 crashes and 6600 near-crashes. The lower-level nest, defined only for near-crash and crash events, will include various incident types such as rear-end, road departure, and turn-into-path (the raw data has more than 16 categories, these will be aggregated into a reasonable number of groups for modeling). Road-departure appears to be dominant incident type for crashes while rear-end and side-swipe are dominant incident types for near-crashes. Data Request Scope Selected variables within the Event detail table, Trip, Participant and Vehicle data sets are being requested. The event detail table holds information about near-crashes, crashes etc. which allows for performing nested structures for in the prediction model. The remaining datasets,Trip, Participant and Vehicle data sets provides additional but important information about the event characteristics, information about persons involved in the event (demographics, cognitive and medical conditions etc.) and lastly the vehicle characteristics pertaining to the event occurrence. Data Specification Trip Summary Table Event Detail Table Participant Driver Demographics Questionnaire Driving History Questionnaire Medical Conditions Medications Barkley's ADHD Screening Test Visual and Cognitive Test Sleep Habit Questionnaire Clock Drawing Assessment Sensation Seeking Scale Risk Taking Questionnaire Driver Behavior Questionnaire Vehicle Detail