This is the final report of a research program entitled "Accident Causation Methodology Development for the National Accident Sampling System". With implementation currently underway of a National Accident Sampling System (NASS) to provide nationally-representative data on traffic accidents, this research was undertaken as one of several steps required in the development of an overall analysis system for NASS. The report explores issues related to the design of statistical methods to estimate the determinants of traffic accidents using NASS data; recommends criteria for selecting pre-crash variables for use in analysis; identifies relevant exposure measures, decision rules for their definition, and data sources; discusses a method of estimating multi-dimensional contingency tables from supplementary margins of varying detail and drawn from different samples which "represent" the same population; and demonstrates analysis procedures for handling some of the peculiar problems of traffic accident data analysis and rate computation. Results of a field trial of a preliminary methodology design are reported, along with recommendations for further development. A basic design for a causation methodology appropriate to NASS is recommended.
Abstract