Modeling Repeated Multinomial Route Choices Under Advanced Traveler Information System: Generalized Estimating Equations with Polytomous Logit Function

Abstract

Correlated multinomial route choice data were modeled under an advanced traveler information system (ATIS). A travel simulator was used as a dynamic data collection tool. The simulator uses realistic network and real historical traffic volumes. It provides five levels of ATIS and accounts for types of delay. A 25-node and 40-link network was used. Sixty-three qualified subjects performed a total of 539 trial days. Forty-four distinct routes were chosen. The multinomial generalized estimating equation (MGEE) methodology was used with a generalized polytomous function and an exchangeable correlation structure. MGEEs account for the serial correlation between repeated choices made by the same subject as well as the correlation due to the overlapping between alternatives. The modeling results show that drivers can identify and follow the shortest path when they are provided with advice-free traffic information on all the network links. In addition, it is shown that the odds of choosing a certain shortest-path route, whether the drivers are advised or not, vary from route to route, depending on the characteristics of the route itself. It was proved that the proposed model could account for a correlation in the multinomial repeated route choices with simple computational effort, even for a large number of alternatives.