Correlation and Factor Analytic Approach to Multidimensional Scaling

Abstract

Multidimensional scaling (MDS) attempts to represent higher dimensional p-variate
vectors in lower dimensional spaces such that the interitem proximities are closely
preserved. The paper suggests two (2) new procedures for performing MDS that
minimizes a stress function using pairwise correlations and by utilizing aspects of factor scores in factor analysis. Results reveal that the two (2) procedures perform as well as, if not better than, the classical minimization search algorithm.