Two-Step Approach for Testing Factorial Invariance: Selecting a Reference Variable and Locating Noninvariant Variable

• In Event: Roundtable Session 46
  In Roundtable Session: Applications of Factor Analysis and Structural Equation Modeling

Abstract

If a chosen reference variable is actually noninvariant in testing factorial invariance, one could draw invalid conclusions regarding the source of noninvariance. Here, we evaluated newly proposed two-step approaches: (1) selecting a reference variable (RV) using the smallest modification index (Min-mod) in either a fully or a partially constrained invariance model and (2) locating noninvariant variable(s) through a Wald test. In the first step, the Min-mod was found to be successful to select a truly invariant RV in both models. In the second step, the two-step approach using the RV selected in a partially constrained invariance showed the highest perfect recovery rates with maintain very low Type I error rates while outperforming the sequential use of the largest modification index.

Authors

• Eunju Jung, Indiana University - Bloomington

• Myeongsun Yoon, Texas A&M University