Canonical correlation and discriminant analysis

Abstract

Suppose we have observed pairs of functions (X i ,Y i ), i = 1,...,N, such as the hip and knee angles for the gait cycles of a number of children, as discussed in Section 6.5. We saw there how we can use principal components analysis to examine the variability in the two sets of curves taken together. In this chapter, we pursue a somewhat different emphasis by considering canonical correlation analysis (CCA), which seeks to investigate which modes of variability in the two sets of curves are most associated with one another. Thus in the gait analysis example, we might ask how variability in the knee angle cycle is related to that in the hip angle.