Common Factor Score Estimates in Multiple Regression Problems

Abstract

The issue of alternative methods for estimating factor scores is raised and, in particular, the case of using common factors as independent variables in regression problems is singled out for investigation. A Monté Carlo simulation experiment (varying: factor scoring method, uniqueness, assumption about factor correlations, and number of observed variables factored) is reported. Though several factor score estimators perform “equivalently” in the simulation, the best overall results are obtained with the Dwyer factor extension technique.