Developments in Principal Component Analysis
Bernard D. Flury · 1995
Abstract Principal Component Analysis is an old statistical technique, and most of the important mathematical results have been known for some 40 or more years. Routine practical applications became possible by the development of good software for extracting eigenvectors and eigenvalues of symmetric matrices. Since the sixties of this century, many interesting applications of principal component analysis have appeared in the literature, most prominently in biology, but also in social sciences. To the pioneers in the development of the theory of principal component analysis (like Pearson 1901, Hotelling 1933, Anderson 1963), it may have been almost unthinkable that people would some day perform analyses with dozens, if not hundreds, of variables. Indeed, the availability of efficient software has not always bene fitted science, because investigators are tempted to let the computer do the thinking. Too many people use principal component analysis (and other multivariate techniques) to study too many variables, hoping that putting ever more data into an existing algorithm will lead to valid scientific conclusions.