Abstract
An axiomatic probabilistic model for the analysis of accident data between classes of drivers in terms of parameters for proneness and exposure of these classes is developed. Statistical goodness of fit criteria and numerical solutions for optimal estimation of the parameters are given.