This thesis takes the statistical approach to road safety issues as a starting point, for example: to what extent do specific developments or measures contribute to the decrease in the number of traffic casualties and what does that teach us for the future? For a quantitative answer to this type of question, the author searches for numerical relations between the observed changes in the number of traffic crashes or the number of traffic casualties on the one hand, and the possible explanations on the other hand: changes in policy, changes in traffic behaviour, changes in exposure, the introduction of specific measures, et cetera. Both the road safety data and the possibly explanatory data show many imperfections and inaccuracies. If this is not taken into account in an analysis it can result in incorrect or unjustified conclusions. This thesis presents a new approach toward the customary time series analyses, specifically from the road safety research angle. The approach is based on so-called structural time series models and combines three fundamental aspects simultaneously. This makes it possible to include multiple explanatory variables in the model as well as multiple dependent variables like the number of accidents and the number of casualties. The model also addresses observational errors or even data that are missing altogether. Finally, the methodology enables the researcher to decompose the road safety process and it uncertainties in interpretable components. Using empirical road safety data the thesis demonstrates the effect and the applicability of the methodology. Topics that are discussed in his thesis are, for example, the registration level of accidents, the ways of determining urban and rural mobility, and the effects of weather conditions on road safety. (Author/publisher)
Abstract