UCLA

A new method is developed for estimating nonmetric multidimensional­scaling solutions from pairwise explosion of rank-order data. The class of MDS methods discussed are adaptations of random-utility choice models to the per­ceptions of dissimilarities and are therefore called choice-theoretic MDS mod­els. We propose simple alternatives to maximum-likelihood estimation of the parameters of these models. The estimation techniques are applied to both simulated data and to a study of perceptions of brands in the Japanese beer market.