Again and Again … Repeated Advice for Not Performing Principal Component Analysis as a Substitute for Factor Analysis

Abstract: It has regularly been noted that principal component analysis is inappropriate in the context of construct validation of scales. Nevertheless, several studies perform construct validation of scales by means of principal component analysis. The reasons for clinging to the myth of similarity of principal component analysis and factor analysis against well-known advice are investigated. Sometimes, factor score indeterminacy is mentioned as an argument against factor analysis and in favor of principal component analysis ( Wilkinson, 1989 ). Sometimes larger principal component loadings than factor loadings may encourage researchers to perform principal component analysis. Therefore, we propose a method to unboost component loadings by estimating factor loadings from published component loadings. Moreover, we show that inflated component loadings lead to inflated reliabilities and to model misfit and that this can have detrimental consequences for item selection. In contrast, reporting indeterminacy of scores resulting from factor analysis is recommended as a basis for careful score interpretation.