AbstractThis 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.
Lehmann, Donald, and William Moore. "A Combined Simply Scalable and Tree-Based Preference Model." Journal of Business Research 22, no. 4 (June 1991): 311-26.