ABSTRACT
The INDSCAL (individual differences multidimensional scaling, Carroll and Chang, 1970) model not only extends the characteristics of the MDS model (see Part One) but also goes beyond it on several major points. INDSCAL is very similar, but not identical, to factor analysis. Like the latter, INDSCAL aims at revealing a structure of individual differences or viewpoints against the background of shared reference points. It describes both a stimulus space and a person space. MDS analyzes a single matrix of distances or similarities. This matrix usually results from the mean of the responses of all the questioned subjects. Inter-individual differences or variations vanished in the computation of such a matrix. INDSCAL, by contrast, analyzes as many matrices of distances (or similarities) as there are individuals in the sample. Our reasoning about factor analysis, i.e., that it aims to describe inter-individual variations sparingly, may also apply to INDSCAL in that the latter aims at describing the variations between different distance matrices.
