A NEW NONMETRIC MULTIDIMENSIONAL-SCALING METHOD FOR SORTING DATA

A multidimensional scaling model to deal with sorting data is develope d. The model assumes (a) that stimuli are embedded in a multidimension al space, and (b) that a particular sorting is generated if and only i f all intra-cluster distances between stimuli in the space are smaller than all inter-cluster distances. A new parameter estimation method i s proposed which yields a configuration that satisfies the above requi rement (b) as much as possible for each sorting given as data. This mo del is extended to the individual differences scaling model. The model s are applied to a set of artificial sorting data and two sets of real data.