Abstract
Methods are described, by the application of which the probabilities of injury to car occupants in collisions of all types may be estimated from data on injury accidents alone, without having to make counts of accidents in which no injury had occurred. To do this, however, the velocity changes have to be known for all vehicles in the sample, so that separate probabilities of injury may be calculated for a number of successive small intervals of velocity change covering the range of velocity change in the accident sample.