ABSTRACT
In this chapter we consider the special case when the table of frequencies is square and the rows and the columns refer to the same set of objects in two different states. Such data are found in many situations, for example social mobility tables, confusion matrices in psychology, brand switching tables in marketing research, cross-citations between journals, transition matrices between behavioural states and migration tables. These tables are often characterized by relatively high values down the diagonal, which is such a strong source of association that the more subtle patterns off the diagonal are not seen in the major principal axes. One approach to applying CA to square tables is to split the analysis into two parts: (i) an analysis of the symmetric part of the table, which absorbs the main component of inertia, including the diagonal, and (ii) an analysis of the remaining part of the table called the skew-symmetric part, which contains the information off the diagonal. It is the visualization of this latter component that shows the amount and direction of “flow” from the rows to the columns and vice versa.
