scientific article; zbMATH DE number 3984364

From MaRDI portal
Publication:3747535





A unified method is proposed for sensitivity analysis in descriptive multivariate methods (including principal component analysis, canonical correlation analysis, correspondence analysis, and Hayashi's third method of quantification, etc.) formulated by the generalized singular value decomposition. In this paper the main objective is to evaluate the influence of a small change of the input data to the result of analysis, to find influential observations and judge whether the result is stable or not. The mathematical technique used is the perturbation theory of eigenvalue and eigenvector problems. Some numerical examples are shown to illustrate the proposed method. \textit{S. Sun} [Statistics. China-Japan Symposium on Statistics held in November 1984 in Beijing, China. Contributed papers, extended abstracts. 263-267 (1984)] studied the similar problem for factor analysis.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3747535)