Safety practitioners are benefiting from an increasing volume of research that quantifies collision prediction so that a single and precise collision frequency can often be calculated. However, decision-makers are often interested in determining two additional qualities of a safety analysis: accuracy, and the worst-case (and/or best-case) scenario. Though the discussion of variability often removes the possibility of a single precise answer, the examination of errors can quantify the confidence associated with an analysis and therefore greatly increase the defensibleness of an analysis and the soundness of any decisions based thereon (particularly in economic applications). Furthermore, where error can be quantified, there is an opportunity for the safety professional to justify departing from the mean predicted values based on their expertise and experience, as well as a deep knowledge of the relevant research and underlying data sets. This paper presents three methods of discussing uncertainty, or variability, in safety analyses. The application of this information in conducting comparison studies of alternative intersection control is highlighted in a case study that demonstrates how the confidence and defensibleness of safety analyses is improved. These methods present a practical continuation to Highway Safety Manual processes and can serve to enhance an agency’s safety policies.
