Abstract
Spatial choice models have been criticized for their restrictive assumptions. In particular, the independence or constant-ratio axiom of Luce has been questioned because of its failure to account for the effects of similarity between alternatives. Tversky's elimination by aspects model provides a natural explanation of similarity effects, but cannot be estimated from binary probabilities. The elimination by tree or, equivalently, hierarchical elimination model is less restrictive than the constant-ratio model and more parsimonious than elimination by aspects. To estimate this model in a spatial interaction context, a hierarchical regionalization based upon interareal flow data is first obtained. The lengths of the links in the resultant tree structure are reestimated on the basis of choice probability ratios.
Get full access to this article
View all access options for this article.