Semiparametric estimation of (constrained) ultrametric trees
Michel Wedel, Wayne S. DeSarbo · 1996
This paper is concerned with the semi-parametric estimation of Ultrametric tree-representations of subjects ' paired comparisons of stimuli, and captures subject heterogeneity using a finite mixture formulation. In many other approaches to the analysis of subjects decision processes, such finite mixture models have been gainfully applied. A new likelihood based estimation methodology is presented for Ultrametric tree structures that accommodates the Ultrametric constraints. This estimation procedure in addition permits the incorporation of a variety of additional external restrictions on the tree structure. Correlations among the observed dissimilarities are allowed for. The performance of the method to identify Ultrametric trees is investigated on synthetic data and an empirical application to published data from Schiffman, Reynolds, and Young (1981) is provided. The ability to deal with specific constraints on the tree-topology is demonstrated.