Car injury site analysis : a Bayesian approach

Abstract

The purpose of this dissertation was to study and apply a statistical method to the data of all drivers injured within the last 56 months. One of the main statistical tools used nowadays, is the Empircal Bayes' method. The Interactive Highway Safety Design Model is estimated using such a method. In this study, we are taking into account, various road sections on which drivers got injured when involved in car accidents. The main aim is to emphasize certain road sections which are highly hazardous to drivers. In conclusion we shall identify these sites, so site improvements can be implemented. Various statistical tools were studied and implemented, namely hypothesis testing, and Bayesian estimation. Maximum Likelihood Estimation is also an essential part of this dissertation as it was used to estimate the parameter of the Multinomial-Dirchlet Distribution. Initially hypothesis testing was implemented on the data, subsequently Bayesian Estimation was used to see if certain sites have a higher probability than others.