Abstract
This study constructs the nested logit model and error component logit model to investigate the alternative similarity across four intercity modes. Model fitness and behavioral implication are evaluating criterion to compare the differences between the both generalized extreme value models. The mixed logit with overlapping common error component could capture correlation across alternatives as well as the nested logit model. Both models all indicated that two railroad modes, rail and high speed rail, are similar alternatives. Specifying individual characteristics as interaction with mode attributes could significantly capture the systematic effect of response heterogeneity. The results prove that the parameterization of variance in error components could effectively reveal the nest structure across alternatives and indicate which factor cause the heterogeneity of alternative similarity.
