Testing Dependent Correlation Coefficients via Structural Equation Modeling

Abstract

Organizational researchers are sometimes interested in testing if independent or dependent correlation coefficients are equal. Olkin and Finn and Steiger proposed several statistical procedures to test dependent correlation coefficients in a single group, whereas meta-analytic procedures can be used to test independent correlation coefficients in two or more groups. Because computer programming is usually involved, applied researchers may find these procedures hard to implement, especially in testing the dependent correlation coefficients. This article suggests using a structural equation modeling (SEM) approach as a unified framework to test independent and dependent correlational hypotheses. To demonstrate the comparability among these approaches, examples and ad hoc simulation studies are used. Advantages of the SEM approach are also discussed.