Quality control techniques were applied to the analysis of traffic accident information. The underlying theory of control charts for accident data is presented. Several new theoretical conclusions relating to the statistical sensitivity of accident control charts were formulated. These same control techniques were also adapted to operational decision-making processes. A generalised system of control chart computer programs, designed to reduce the reported statistical techniques to operational practice, was developed and subsequently applied to a sample of accident data. Results suggest that for many accident data uses, a high alpha-error probability should be tolerated in order to realise a low beta-error probability concomitant with operationally defined lengths of roadway, realistic control chart time periods, and reasonable sensitivity to changes in accident producing potential. Such control charts would rarely fail to detect a small change in accident potential at the cost of having many of these change indications be spurious. The principle is that for many accident data applications, it is appropriate to tolerate many false indications of change in order to reduce the likelihood of failing to detect a real change. (Author/publisher)
Abstract