Applying Components Analysis to Attitudinal Segmentation
Principal components analysis (PCA) is well known in both the quantitative psychology and marketing research communities. However, there is notable confusion of PCA with factor analysis (FA) that persists both in the literature and in practice. While graduate training plays amajor role in this confusion, the situation is exacerbated by the use of computer programs with options for both PCA and FA (e.g., PROC FACTOR in SAS software). Users who run such programs without changing the default options are often unaware that they might have carried out a PCA while intending to do FA (defaults in most major statistical packages make it unlikely that FA will be carried out instead of PCA.)
Browne and Tateneni (2009) examine the similarities and differences between FA and PCA by considering the dominant purpose of each of these methods of data analysis. The dominant purpose of a method of data analysis is determined by the assumptions made in the derivation of the method, not by common usage (or misuse). Rethinking PCA in terms of dominant purpose leads us away from
the restrictive identiﬁcation conditions of principal components to thinking more generally about “optimal components.” This approach also enables the rotation of principal components, which destroys the identiﬁcation conditions of PCA, but retains optimality in the sense of predicting the original variables from a small set of “component” variables.