A new form of statistical model for pedestrian accidents is given. It is tested on data obtained from 55 sites with signalled pedestrian crossings in Edinburgh. The analysis covers three types of road section, including the signalled crossings. The new models for the different types of section have more unity than do the corresponding log-linear models. They predict that, at low pedestrian flows, the risk is the same for the three types of section. In comparison with the common log-linear models for the same data, the new models have better fits; they do not, however, cover high pedestrian flows. In addition to section type, the only explanatory variables in either form of model are pedestrian flow and road width. It is found that the risk in sections affected by pedestrian refuge is greater than in sections not so affected. This appears to contradict some other findings, but it might apply only to refuges near to signalled crossings. (Author/publisher)
Abstract