Multivariate statistical analysis for exploring road crash-related factors in the Franche-Comté region of France

Abstract Understanding and modeling road crash data is crucial in fulfilling safety goals by helping national authorities to take necessary measures to reduce crash frequency and severity. This work aims at giving a multivariate statistical analysis of road crash data from the French region of Franche-Comté with special attention to road crash gravity. The first step for this multivariate analysis was to perform multiple correspondence analysis in order to assess associations between the road crash injury and several important accident-related factors and circumstances. Log-linear models are used next in order to detect associations between road crash severity and related factors such as alcohol/drug consumption or spatial crash locations. The effects of each factors have been also evaluated on the road crash gravity by using ordinal logistic regression. Data used in this study are extracted from BAAC files, the French census of road crashes.