A Simulation Study of Factor Score Indeterminacy

Abstract

Though factor analysis continues to be one of the most frequently used multivariate techniques, its value has been questioned because of the indeterminacy of factor scores. The authors review the literature on factor score indeterminacy and discuss the implications of indeterminacy for research practice. A simulation experiment is used to investigate the effects of characteristics of the factor analytic data set on the accuracy of factor score estimation. Indeterminacy is found to depend critically on the level of communality and to be detected more accurately via image factoring than by principal axis or principal component analysis.