A nonparametric quasi-experimental time series analysis is presented as a method for circumventing many of the problems often encountered in applying statistical analysis to the evaluation of safety programs. It is shown how this method is applicable where classical experimental techniques are impossible to implement. A case study, which shows the application to an actual pedestrian program evaluation, is given to exemplify the technique.

Under ideal conditions, experiments are set up in such a manner that by varying certain factors under controlled conditions, observation can be made as to the effects of the factors involved. The manner in which these factors are combined, the sequence in which they are observed, and the conditions under which they are controlled constitute the experimental design. Unfortunately, not all situations that need evaluating lend themselves to the controls that are dictated by classical experimental design. It is not always possible to randomly assign treatment conditions or to control unknown factors. Under these circumstances, a" quasi-experimental" design may be considered.