This chapter presents multidimensional scaling (MDS) methods and their application to customer satisfaction surveys. MDS methods are multivariate statistical analysis techniques of particular relevance to survey data analysis. In fact, despite some criticism, such applications are gaining in popularity, especially in market research studies. The chapter begins by presenting the theory of MDS, including theoretical results on so-called proximity data, the basic input data of MDS. An overview is given of the most widely applied MDS models, the classical, least squares and nonmetric MDS. Several reserach topics in MDS are also considered, i.e. the problems of assessing goodness of fit, comparing two different MDS solutions, and diagnosing anomalous results that could derive from analyses. The chapter then goes on to deal with the application of metric MDS models to the ABC annual customer satisfaction survey. It concludes by outlining some future directions for MDS research.
18.1 An overview of multidimensional scaling techniques
Multidimensional scaling (MDS) is the name given to a large family of multivariate data analysis techniques for dealing with dimensionality reduction and data visualization problems in situations where the reproduction of ‘closeness’ of data from an observational space is of primary concern. The common, basic idea of these methods is to represent a given set of proximities, that is, measures of pairwise similarity/dissimilarity ...