Statistical Aspects of Injury Severity Part IV: Matched Data

By matched data is meant information about two casualties in the same accident (or in very similar accidents that have been matched from a large data set). The two casualties might differ in, for instance, which vehicle they were in, or whether they were or were not wearing a seat belt. Two strategies for the statistical analysis of such data are described. One explicitly models the probability of each combination of severities of injury. The second concentrates on a parameter describing how different are the severities to the two classes of casualty. Several data sets from two fields of application are presented and analyzed as matched data. The first field is the effect of mass ratio on the severities of injury to drivers of vehicles involved in head-on collisions. The second is the difference in severity between belted and unbelted car occupants. The same methods are applicable, of course, to the analysis of other factors. Finally, as an example of the extensions possible with such models, it is shown how it may be feasible to make a distinction between the effects on injury severity of vehicle design and of impact speed, without actually knowing the impact speeds of the accidents.