Comparison of analytical approximation methods for multinomial probit model: A case study with passenger’s rail-route choice analysis in the Tokyo Metropolitan Area
Since multinomial probit model (MNP) can appropriately capture the correlations among overlapped routes in the highly dense network, it is suitable for rail-route choice model in metropolitan area like Tokyo (Yai et al., 1997). However, multidimensional integral included in MNP cannot be computed analytically. Therefore approximation with Monte Carlo Simulation has played a pivotal role. Towards the enhancement of demand forecasting it would become more important to consider taste variations of travelers leading to more runs of probit probabilities in practice. The further speed up of computing MNP thus would be helpful for practitioners. This study examines MNP with four different analytical approximations for rail-route choice applications and demonstrates to compute route choice probabilities in transit assignment for the network of Tokyo Metropolitan Area. Through small and large scale tests we find that there would be enough potential of analytical approximation.