ABSTRACT
Statistical models have been developed to analyze different types of collisions at intersections, and on sections on urban roads, rural roads, carriageways, and motorways. Empirical tests about the relationship between collisions and independent variables (including traffic flows) can be based on different statistical techniques, all having their own limitations. Besides, different data definitions (such as time and spatial units) are used in different applications in the literature; careful definition and selection of the data are rare. Collision data consist of counts and thus are considered as Poisson or negative binomial distribution. This chapter explores the nature of the underlying distribution and methods, which are fundamentally linked to traffic volume, notably average annual daily traffic (AADT). There has been much discussion concerning the aggregation of road segments and the application of Poisson-based methods to analyze larger segments of road collisions. Poisson regression is a commonly used technique; however, research has shown that it is the most effective with lower levels of aggregation.
