The statistical analysis of vehicle crash accident data is generally problematic. Data from commonly used sources is almost never without error and complete. Consequently, many analyses are contaminated with modeling and system identification errors. In some cases the effect of influential factors such as crash severity (the most significant component being speed) driver behavior prior to the crash, etc. on vehicle and occupant outcome is not adequately addressed.
The speed that the vehicle is traveling at the initiation of a crash is a significant contributor to occupant risk. Not incorporating it may make an accident analysis irrelevant; however, despite its importance this information is not included in many of the commonly used crash data bases, such as the Fatality Analysis Reporting System (FARS). Missing speed information can result in potential errors propagating throughout the analysis, unless a method is developed to account for the missing information.
In this paper, a method is presented to estimate the effect of missing speed information on the statistical effect of different variables (age, vehicle mass, …) on injury or fatality. The method developed uses historical data, such as the crash velocity distributions and the injury as a function of velocity -obtained from the National Accident Sampling System/Crash Data System (NASS/ CDS)- to estimate the modeling errors or the confounding effect expected from not considering the velocity factor in accident analysis. The method can be used to correct the level of statistical significance or to adjust the potential effect of explanatory crash variables, obtained from an accident analysis, when there is missing velocity information.
Examples are provided where the suggested analytical procedure is used to estimate the confidence intervals for previously published risk of fatality based on crash data as well as to estimate a factor effect adjustment due to the missing speed data.