The view adopted in this paper is that the basic level for causal modeling in road safety is that of the individual accident, where human decisions interact with physical, deterministic processes. In some situations, statistical analyses of aggregated data can illuminate what happens at this basic level, but if not carefully done, aggregated analyses can also mistake methodological artifacts, such as those arising in Simpson's paradox or the ecological fallacy, for substantive effects. Here the author indcates that this is a feasible program, by identifying a measure of causal effectiveness, called probability of necessity (PN), and showing how it arises at both the basic and aggregated levels of analysis. A connection between PN and accident reduction factors will first be made and then, using a structural causal model, shows how to use PN to assess the potential safety impact of speed controls on individual residential streets. Finally, how PN generalizes the notion of causal factor used in accident reconstruction is illustrated. (Author/publisher) For the covering entry of this conference, please see ITRD abstract No. E208120.
Samenvatting