A Comparison of Ten Methods for Determining the Number of Factors in Exploratory Factor Analysis

Abstract

The effectiveness of 10 methods for estimating the number of factors were compared. These methods were the minimum average partial procedure, the likelihood ratio test, the Akaike information criteria (AIC), the Schwarz information criteria, the common factor and principal component versions of parallel analysis, the standard error scree test, and the eigenvalues greater than average criterion. Two simulation studies were conducted. In the first study, the true number of factors, variable-to-factor ratio, level of communality, and sample size were manipulated. The second study included factor correlations as an additional experimental variable. Common factor parallel analysis was the most consistently accurate method across conditions and studies. Neither eigenvalue rule performed well, but the common factor version was superior. Over almost all conditions, the AIC was either correct or overfactored by one and might; thus, be a reasonable alternative.