The statistical accuracy of Delta-V in systematic field accident studies

Much attention has been paid to the importance of calculating delta-V (the change of velocity during impact) and other impact severity measures accurately. However delta-V cannot be evaluated in every case sampled by a systematic study of road accidents. This can lead to statistical distortions if the subsample of cases for which delta-V is calculated is not representative of the whole sample. This problem has received less recognition than the problem of calculating delta-V accurately when it is calculated. This paper contains new results on the accuracy of CRASH3 delta-V for European passenger cars and a discussion of the problem of calculating delta-V for a representative subsample. On the data available, CRASH3 underestimates delta-V for rigid and deformable barrier impacts but not frontal car-to-car impacts. The statistical results obtained using any single method for calculating delta-V are likely to be unrepresentative of certain classes of impact type, impact severity and collision partner. A flexible approach towards using a variety of methods to evaluate delta-V is necessary to counter this difficulty.