Scholion.ai

Sensitivity of Principal Components

W. J. Krzanowski

Journal of the Royal Statistical Society: Series B (Statistical Methodology) · 1984 · Volume 46 · Issue 3

DOI: 10.1111/j.2517-6161.1984.tb01325.x

Abstract

Simple analytical expressions are derived for the maximum changes in the coefficients of any principal component that are associated with a given (small) change in the variance of the component. This enables a sensitivity analysis to be conducted simultaneously with any principal component analysis, to investigate the stability of the derived components. This has implications for the interpretation of any set of components. A brief illustration is provided, and possible extensions to other multivariate techniques are outlined.

Keywords

eigenvalues, eigenvectors, principal components, response surface

Enable JavaScript for citation tools and additional paper interactions.