In this paper the statistical and predictive performance of two disaggregate choice models that incorporate probabilistic choice setformation are compared with a standard logit specification. The empirical work is conducted with work mode choice data from a Braziliancity. For the type of travel demand analyzed it is found that, although statistically inferior to the probabilistic choice set specifications, the standard logit specification, allied with market segmentation, is a robust formulation in both statistical and predictive terms. Recommendations for future research work in probabilistic choice set modeling are presented.(A).
Samenvatting