THE ASSESSMENT OF CHANGES IN ACCIDENT FREQUENCIES ON LINK SEGMENTS: A COMPARISON OF FOUR METHODS

If sites are selected for accident remedial treatment at least partly on the basis of a high observed accident frequency in a particular time period, then the observed accident frequency is likely to fall after treatment, even if the treatment has had no effect. This, so-called 'regression-to-mean' effect leads to over-estimates of treatment effectiveness. A number of methods have been proposed for estimating the magnitude of the regression-to-mean effect. This paper analyses the performance of three such methods at link sites.