Estimation and Uses of the Proportion of Method Variance for Multitrait-Multimethod Data

Abstract

Recent evidence supports the use of the correlated uniqueness model over the general confirmatory factor analysis model for multitrait-multimethod (MTMM) data The former provides no method factor loadings and therefore no obvious estimate of the proportion of method variance. This is problematic; researchers have used proportions of method variance from the general confirmatory factor analysis model for many purposes. Demonstrations showed that correlated uniquenesses can be averaged to estimate the proportion of method variance. Unlike factor loadings, correlated uniquenesses need not be squared. The first demonstration included a mathematical proof and analysis of an artificial matrix The second involved two employment interview matrixes and one performance appraisal matrix Proportion of method variance estimates were used to make comparisons within and between MTMM matrixes. Results suggested that more structured employment interviews contain smaller proportions of method variance and that peer performance ratings contain greater proportions of method variance than do self-ratings.