SWOV Catalogus


I 892221 /81 / IRRD 892221
Pardillo-Mayora, J.M.
Routes/Roads. 1997/04. (294) Pp5-19 (9 Refs.)

Samenvatting This paper discusses the 'regression-to-mean effect', a very important phenomenon which was discovered by Galton in 1898, and proposes an interesting way of solving this persistent problem in relation to accident sites, which is at present being tested in Spain. For accident statistics, the phenomenon consists of mean accident rates, at sites with the highest numbers of accidents over a certain time period, tending to decrease and approach the overall mean value in successive periods. This is because the accident rates at a given site fluctuate, so that there is a continuous change in which sites have most accidents. Studies in different countries, including Spain, show that regression to mean may generate apparent reductions of 5-30% in the numbers of accidents at sites usually selected for safety improvements. Methods for estimating a site's inherent accident rate can be divided into classical methods of experimental design and Bayesian analyses. A Bayesian method has been developed in Spain for estimating long-term mean accident rates at each site in a road network; it aims to estimate the probability distribution of the inherent accident rate (IAR) of each site. This allows systematic detection of accident black spots, priority ranking of schemes for treating and preventing them, and assessment of the effectiveness of implemented schemes.
Suggestie? Neem contact op met de SWOV bibliotheek voor uw opmerkingen
Copyright © SWOV | Juridisch voorbehoud | Contact