Eigenvector rotation

Fig. 6. Eigenvector rotation. The first eigenvector spans the direction of greatest variance in the data, and x represent the two original variables.

So, according to the aim of data analysis, one method may be selected, or another data transform may be used before the eigenvector rotation. 

Graphical representation is also important to visualize the results attained by the methods that extracted the factors. The objective is not to improve the fit between the observed and reproduced correlation matrices but as an aid to the interpretation of scientific results, making them more understandable. Eigenvector rotation is the most commonly used, and the four types of rotation are as follows ... 

We next solve the secular equation F — I = 0 to obtain the eigenvalues and eigenvectors o the matrix F. This step is usually performed using matrix diagonalisation, as outlined ii Section 1.10.3. If the Hessian is defined in terms of Cartesian coordinates then six of thes( eigenvalues will be zero as they correspond to translational and rotational motion of th( entire system. The frequency of each normal mode is then calculated from the eigenvalue using the relationship ... 

It turns out that the htppropriate X matrix" of the eigenvectors of A rotates the axes 7t/4 so that they coincide with the principle axes of the ellipse. The ellipse itself is unchanged, but in the new coordinate system the equation no longer has a mixed term. The matrix A has been diagonalized. Choice of the coordinate system has no influence on the physics of the siLuatiun. so wc choose the simple coordinate system in preference to the complicated one. 

In the vector space L defined over the field of real numbers, every operator acting on L does not necessarily have eigenvalues and eigenvectors. Thus for the operation of 7t/2 rotation on a two-dimensional vector space of (real) position vectors, the operator has no eigenvectors since there is no non-zero vector in this space which transforms into a real multiple of itself. However, if L is a vector space over the field of complex numbers, every operator on L has eigenvectors. The number of eigenvalues is equal to the dimension of the space L. The set of eigenvalues of an operator is called its spectrum. 

Note that since SVD is based on eigenvector decompositions of cross-product matrices, this algorithm gives equivalent results as the Jacobi rotation when the sample covariance matrix C is used. This means that SVD will not allow a robust PCA solution however, for Jacobi rotation a robust estimation of the covariance matrix can be used. 

Calculation of eigenvectors requires an iterative procedure. The traditional method for the calculation of eigenvectors is Jacobi rotation (Section 3.6.2). Another method—easy to program—is the NIPALS algorithm (Section 3.6.4). In most software products, singular value decomposition (SVD), see Sections A.2.7 and 3.6.3, is applied. The example in Figure A.2.7 can be performed in R as follows ... 

Rotation of the translated factor axes is an eigenvalue-eigenvector problem, the complete discussion of which is beyond the scope of this presentation. It may be shown that there exists a set of rotated factor axes such that the off-diagonal terms of the resulting S matrix are equal to zero (the indicates rotation) that is, in the translated and rotated coordinate system, there are no interaction terms. The relationship between the rotated coordinate system and the translated coordinate system centered at the stationary point is given by... 

---