ABSTRACT
As is observed in Figure 1.4b (Section 1.3), factorial structures of rows and columns of a data matrix do not necessarily emerge as rectangular configurations in CPCA. This is because the row design matrices G j ( j = 1, · · · ,J) representing the factorial structures are put together into a single row block matrixG= [G1, · · · ,GJ ] in CPCA. The column design matrices H j ( j = 1, · · · ,J) are similar. Consequently, the dimensions extracted from CPCA do not correspond one-to-one with any particular G j. To obtain such dimensions, we have to use DCDD (different constraints on different dimensions; Takane et al. 1995), the topic of this chapter.
