Regression analysis has been used in the study of road traffic safety to describe the influence of the traffic volume (number of vehicle kilometers travelled) on the number of accidents. The purpose of this paper is to carry through an analysis based on accidents with low frequencies. In particular the effect of speed limits on fatal accidents and accidents involving severe personal injuries has been studied. The analysis uses the Poisson regression model given by Erlander-Gustavsson-Svensson ("On Asymptotic Simultaneous Confidence Regions for Regression Planes in a Poisson Model", Int. Stat. Rev. Vol. 40, No. 2, 1972). Both maximum likelihood estimates and least square estimates are used. Since most of the results are asymptotic a small simulation study has been made in order to illuminate the problem when the results are applicable. Some comments are also given of the possibility to use a normal regression model as an approximation. (Author/publisher)
Abstract