In statistics, canonical correlation analysis, introduced by Harold Hotelling, is a way of making sense of cross-covariance matrices.

Definition

Given two column vectors X = (x_1, \dots, x_n)' and Y = (y_1, \dots, y_m)' of random variables with finite second moments, one may define the cross-covariance \Sigma _{12} = \operatorname{cov}(X, Y) to be the n \times m matrix whose (i, j) entry is the covariance \operatorname{cov}(x_i, y_j).

Canonical correlation analysis seeks vectors a and b such that the random variables a' X and b' Y maximize the correlation \rho = \operatorname{cor}(a' X, b' Y). The random variables U = a' X and V = b' Y are the first pair of canonical variables. Then one seeks vectors maximizing the same correlation subject to the constraint that they are to be uncorrelated with the first pair of canonical variables; this gives the second pair of canonical variables. This procedure continues \min\{m,n\} times.

Computation

Proof

Let \Sigma _{11} = \operatorname{cov}(X, X) and \Sigma _{22} = \operatorname{cov}(Y, Y). The parameter to maximize is

\rho = \frac{a' \Sigma _{12} b}{\sqrt{a' \Sigma _{11} a} \sqrt{b' \Sigma _{22} b}}.

The first step is to define a change of basis and define

c = \Sigma _{11} ^{1/2} a,

d = \Sigma _{22} ^{1/2} b.

And thus we have

\rho = \frac{c' \Sigma _{11} ^{-1/2} \Sigma _{12} \Sigma _{22} ^{-1/2} d}{\sqrt{c' c} \sqrt{d' d}}.

By the Cauchy-Schwarz inequality, we have

c' \Sigma _{11} ^{-1/2} \Sigma _{12} \Sigma _{22} ^{-1/2} d \leq \left(c' \Sigma _{11} ^{-1/2} \Sigma _{12} \Sigma _{22} ^{-1/2} \Sigma _{22} ^{-1/2} \Sigma _{21} \Sigma _{11} ^{-1/2} c \right)^{1/2} \left(d' d \right)^{1/2},

\rho \leq \frac{\left(c' \Sigma _{11} ^{-1/2} \Sigma _{12} \Sigma _{22} ^{-1/2} \Sigma _{22} ^{-1/2} \Sigma _{21} \Sigma _{11} ^{-1/2} c \right)^{1/2}}{\left(c' c \right)^{1/2}}.

There is equality if the vectors d and \Sigma _{22} ^{-1/2} \Sigma _{21} \Sigma _{11} ^{-1/2} c are colinear. In addition, the maximum of correlation is attained if c is the eigenvector with the maximum eigenvalue for the matrix \Sigma _{11} ^{-1/2} \Sigma _{12} \Sigma _{22} ^{-1} \Sigma _{21} \Sigma _{11} ^{-1/2} (see Rayleigh quotient). The subsequent pairs are found by using eigenvalues of decreasing magnitudes. Orthogonality is guaranteed by the symmetry of the correlation matrices.

Solution

The solution is therefore:

• c is an eigenvector of \Sigma _{11} ^{-1/2} \Sigma _{12} \Sigma _{22} ^{-1} \Sigma _{21} \Sigma _{11} ^{-1/2}
• d is proportional to \Sigma _{22} ^{-1/2} \Sigma _{21} \Sigma _{11} ^{-1/2} c

Reciprocally, there is also:

• d is an eigenvector of \Sigma _{22} ^{-1/2} \Sigma _{21} \Sigma _{11} ^{-1} \Sigma _{12} \Sigma _{22} ^{-1/2}
• c is proportional to \Sigma _{11} ^{-1/2} \Sigma _{12} \Sigma _{22} ^{-1/2} d

Reversing the change of coordinates, we have that

• a is an eigenvector of \Sigma _{11} ^{-1} \Sigma _{12} \Sigma _{22} ^{-1} \Sigma _{21}
• b is an eigenvector of \Sigma _{22} ^{-1} \Sigma _{21} \Sigma _{11} ^{-1} \Sigma _{12}
• a is proportional to \Sigma _{11} ^{-1} \Sigma _{12} b
• b is proportional to \Sigma _{22} ^{-1} \Sigma _{21} a

The canonical variables are defined by:

U = c' \Sigma _{11} ^{-1/2} X = a' X

V = d' \Sigma _{22} ^{-1/2} Y = b' Y

Hypothesis testing

Each row can be tested for significance with the following method. If we have p independent observations in a sample and \widehat{\rho}_i is the estimated correlation for i = 1,\dots, \min\{m,n\}. For the ith row, the test statistic is:

\chi ^2 = - \left( p - 1 - \frac{1}{2}(m + n + 1)\right) \ln \prod _ {j = i} ^p (1 - \widehat{\rho}_j^2),

which is asymptotically distributed as a chi-square with (m - i + 1)(n - i + 1) degrees of freedom for large p.^[1]

Practical uses

A typical use for canonical correlation in the psychological context is to take a two sets of variables and see what is common amongst the two sets. For example you could take two well established multidimensional personality tests such as the MMPI and the NEO. By seeing how the MMPI factors relate to the NEO factors, you could gain insight into what dimensions were common between the tests and how much variance was shared. For example you might find that an extraversion or neuroticism dimension accounted for a substantial amount of shared variance between the two tests.

References and links

• See also generalized canonical correlation.
• Applied Multivariate Statistical Analysis, Fifth Edition, Richard Johnson and Dean Wichern
• Canonical correlation analysis - An overview with application to learning methods http://eprints.ecs.soton.ac.uk/9225/01/tech_report03.pdf, pages 5-9 give a good introduction Neural Computation (2004) version
• FactoMineR (free exploratory multivariate data analysis software linked to R)

pl:Korelacja kanoniczna sv:Kanonisk korrelationsanalys

---

Source: Wikipedia | The above article is available under the GNU FDL. | Edit this article