With emphasis on a newly finished traffic restraint programme carried out in Österbro, Copenhagen (see paper by Engels, IRRD 265001) the mathematical statistical tools for evaluation are discussed. The exposition will fall in two parts. The first will deal with the statistical procedures related to the actual accident analysis. The second will be the analysis of different behavioural studies of the street-users with emphasis on speed-measurements of motorized vehicles. The statistical tools for the accident analysis have two main purposes. The first is: "has the accident rate actually changed?" while the second is "in case of a change, what is its magnitude and what are its limits?" the first question is answered by means of statistical tests using multiple-dimensional log-linear Poisson models. A special case of control-area (placebo) philosophy is used here. The second question about the size of the effect is answered by more conventional statistical theory but carried out as well with as without a control-area. Supporting the accident-analysis, behavioural studies have been carried out. The speed measurements are dealt with in mathematical statistical terms using analysis of variance, investigating the assumptions of these analyses and dealing with general behaviour of street-users in terms of log-linear Poisson models. Finally some conflict and accident studies by Zimolong (1980) are reanalyzed by means of log-linear Poisson models.(a) for the covering abstract of the conference see IRRD abstract no 264967.
Samenvatting