An Improved Procedure and Program for Minimum Residual Factor Analysis

Abstract

An improved iteration procedure is offered for minimum residual factor analysis, a method which requires no communality estimates. Successive iterated vectors are length adjusted instead of averaged. Length adjustment requires fewer iterations to achieve convergence and stops the factoring process at a more interpretable point, namely, where two successive vectors are equal but opposite in sign. Empirical comparisons are made between length adjustment and averaging methods of iteration. Empirical comparisons are made between the minimum residual method and the principal factor method. Minimum residual factor analysis may offer a better criterion for terminating factor extraction than those in current use.