A COMBINED SIMPLY SCALABLE AND TREE-BASED PREFERENCE MODEL

This article proposes a model that nests both a strict tree model and the Luce choice model. The multiplicative formulation allows for easy estimation using least-squares procedures. The model is shown to be more parsimonious than the hierarchical elimination method and in a small illustration, to significantly out-perform Luce in predicting soft-drink preferences. © 1991.