Mixture Unfolding Models

We describe a general mixture unfolding approach that allows simultaneously for a probabilistic classification of observations into segments (similar to the GLIMMIX models described in the preceding chapter) and the estimation of an internal unfolding model within each segment. This multidimensional scaling (MDS) based methodology is formulated in the framework of the exponential family of distributions, whereby a wide range of data types can be analyzed. Possible re-parameterizations of stimulus coordinates by stimulus characteristics, as well as of probabilities of segment membership by subject background variables, are permitted. We also review previous applications of the approach to market segmentation problems.