Identification of sites for road accident remedial work by Bayesian statistical methods: An example of uncertain inference

Road accident remedial work can be effected in many cases by implementing low cost engineering measures. These are intended to restore the safety performance of sites to which they are applied to prevailing levels according to their kind, traffic flows and design. This approach is therefore appropriate to sites where the mean accident frequency exceeds a standard value for that site. In this paper, we present analyses that provide quantitative indications of which sites should be considered for this kind of treatment according to their accident record. We describe four measures that are informative in this respect - the count of accidents, an estimate of the mean accident frequency, the amount by which this estimate exceeds a standard value, and the probability that it does so. We illustrate the use of this analysis by applying it to a dataset of accident records of priority controlled road junctions and compare the results of the different analyses.