Transfer and updating of Logit models of gap-acceptance and their operational implications

Abstract

This paper deals with the problem of model transferability in the specific case of Logit models of gap-acceptance behavior. Transfer Scaling, Bayesian Updating and Combined Transfer Estimation are adopted as transfer methods, and the effectiveness of model transfer is evaluated, on the basis of experimental data collected at two priority intersections, using several indicators proposed in the travel demand modeling literature. The main conclusions are that the accuracy of the transferred models is generally similar to that of the locally estimated ones, and that the method known as Combined Transfer Estimation performs best among the tested approaches. These results are not significantly affected by the size of the sample of observations used for model transfer. Moreover, the potential impacts of the differences between transferred and locally estimated models on capacity estimates appear to be negligible for practical purposes.