عنوان مقاله [English]
Modeling behavior of individuals in destination choice plays an important role
in travel demand analysis and hence assessing transportation policies impact.
Most researchers have used multinomial logit model, which does not allow for
heterogeneity among individuals and alternatives. However, mixed logit models
account for random distribution of coefficients in the utility function, hence
heterogeneity of individuals. Heteroscedastic extreme value model, on the other
hand, assumes an independent and non-uniform distribution of the error term,
which leads to heterogeneity among alternatives. The purpose of this paper is
to examine heterogeneity among individuals and alternatives in shopping trips
destination choice in Qazvin. Multinomial logit (as the base model, not
considering heterogeneity), mixed logit and heteroscedastic extreme value are
calibrated and their results are compared. Unlike previous studies, individuals
choose destination based on distance to their residence. Model results of 1570
shopping O-D trips based on various criteria (such as log likelihood and goodness of fit coefficients) indicate superiority of the mixed logit model. It was also observed that the individuals behave heterogeneously considering distance and departure time in choosing their shopping destinations. On the other hand, heteroscedastic extreme value model results were not much better than multinomial logit, thus relaxing the assumption of homoscedasticity if the error term in multinomial logit may not be necessary considering the small increase in model fit. This paper is an extension to conventional destination choice models by taking account of heterogeneity among individuals and alternatives enabling more realistic results helping decision makers in setting more effective policies.