Shift, slope and curvature for a class of yields correlation matrices (Q996318)

To develop statistical models that describe the movement of yield curves and to validate models for interest dynamics, principal component analysis is used. The principal components are the leading eigenvectors of a correlation matrix. This matrix is symmetric and may or may not have certain properties for its column entries. The exponential model leads to studies of correlation matrices \(R\) of type \[ \left( \begin{matrix} 1&\rho&\rho^2& \dots & \rho^{n-1}\\ \rho & 1 & \rho & \dots & \rho^{n-2}\\ \vdots& \vdots&\vdots&\ddots&\vdots\\ \rho^{n-1} & \dots &&\dots & 1 \end{matrix} \right) \] for \(0 < \rho \;< 1\). The main results of this paper deal with the sign distribution of the leading three eigenvectors of \(R\). The proofs use Perron-Frobenius, interlacing, and the power method and compare the discrete results with those for the continuous exponential model.