Before-after studies in road safety are not statistical experiments since treatment is almost never applied to entities selected at random. Therefore the interpretation of before-after studies requires both greater care and also ingenuity. This book provides guidance both on "how to do it" and on "what it means". The first is important for the research-analyst who has to design and interpret a before-after study, the second to the transportation professional who needs to form an opinion about the conclusions of a report. The book is divided into three parts. The first contains information that is essential for the understanding and conduct of a before-after study. The second part provides an adaptation of conventional approaches, such as before-after with comparison group, to the realities of observational studies. The third part is devoted to an exposition of new approaches to the interpretation of observational before-after studies. Multivariable models are coupled with an Empirical Bayes framework to furnish a coherent and powerful tool. Numerous numerical examples and case studies are provided.
Abstract