Multitrait-Multimethod Comparisons Across Populations: A Confirmatory Factor Analytic Approach.

Area of Science:

• Psychometrics
• Quantitative Psychology
• Structural Equation Modeling

Background:

• Traditional methods like MANOVA assume measurement equivalence across populations, which may not hold.
• Multitrait-multimethod (MTMM) methodology provides a framework for assessing construct validity and reliability.
• Maximum likelihood confirmatory factor analysis (ML-CFA) is a powerful tool for modeling complex relationships.

Purpose of the Study:

• To present an integrated framework combining MTMM and ML-CFA for cross-population comparisons.
• To enable rigorous statistical analysis of measurement equivalence and construct validity.
• To detect true score-observed score regression intercept and true mean differences between populations.

Main Methods:

• Employs a sequence of hierarchically nested confirmatory factor analytic models.
• Utilizes maximum likelihood estimation for model fitting.
• Compares the proposed procedure with traditional MANOVA using an artificial data set.

Main Results:

• The proposed framework allows for statistical testing of measurement equivalence and construct validity assumptions.
• It differentiates between true score and observed score differences, and true mean differences.
• Demonstrates potential for spurious results with MANOVA versus veridical results with the proposed method.

Conclusions:

• The integrated MTMM-ML-CFA framework offers a statistically sound approach for cross-population comparisons.
• It addresses limitations of MANOVA by explicitly testing measurement assumptions.
• This methodology enhances the validity of cross-group comparisons in psychological research.