Comparing Classic and Novel Approaches to Measurement Invariance

Magraw-Mickelson, Zoe; Hermida Carrillo, Alejandro; Weerabangsa, Maas Misha'ari; Owuamalam, Chuma Kevin; Gollwitzer, Mario

Abstract

Measurement invariance (MI) is vital to any comparison of heterogeneous groups. With multiple-group confirmatory factor analysis (MG-CFA), which is the standard practice for testing MI, there are widely acknowledged limitations, especially with a large number of groups for which strict invariance is difficult or impossible to achieve. New methods, specifically Alignment optimization, gives increased flexibility and new opportunities to make comparisons across a large number of groups. This article compares MG-CFA with Alignment optimization for MI testing in three demonstrative studies. First, in a study of eight countries looking at eight different measures, the MG-CFA method found strict MI is problematic; only partial invariance was achievable and additionally required some of the countries to be excluded in some analysis. However, in using the Alignment optimization method there were not these drawbacks. Next, in a cross-validation reanalysis of existing data from five countries, we tested the reproducibility of conclusions from Alignment and MG-CFA analyses, finding only the results of the Alignment method reproducible. Finally, we show how to apply a recent extension of the Alignment optimization method, Alignment-within-CFA (AwC), to correlational analysis with samples from 30 countries. In a comparison of results with and without Alignment adjustments, we found some differences in conclusions drawn. This article illustrates the differences between methods as well as how to apply them to cross-cultural research.

Citation

Magraw-Mickelson, Z., Hermida Carrillo, A., Weerabangsa, M. M., Owuamalam, C. K., & Gollwitzer, M. (2021). Comparing Classic and Novel Approaches to Measurement Invariance

• Working Paper Type: Working Paper
• Publication Date: Jan 10, 2021
• Deposit Date: Dec 7, 2023
• Public URL: https://durham-repository.worktribe.com/output/1984614
• Related Public URLs: https://doi.org/10.31234/osf.io/pz8u9