An iterative strategy for referent variable selection in MIMIC-interaction modeling.

Area of Science:

• Psychometrics
• Statistical Modeling
• Cross-Cultural Psychology

Background:

• Measurement invariance testing is crucial for examining cross-group latent differences.
• Selecting a referent observed variable is a key limitation in multiple-group confirmatory factor analysis (MGCFA).
• Existing repetitive methods for referent variable selection in MIMIC-interaction models have limitations.

Purpose of the Study:

• To propose and evaluate an iterative search strategy for referent variable selection in MIMIC-interaction models.
• To enhance the accuracy of referent variable identification, particularly under conditions of increased noninvariance.
• To compare the performance of the iterative strategy against the repetitive strategy.

Main Methods:

• Employed a Monte Carlo simulation design.
• Investigated the effects of proportion of noninvariance, magnitude of noninvariance, latent variable differences, and sample size.
• Assessed performance using the accuracy rate of selecting truly invariant observed variables.

Main Results:

• The proposed iterative strategy generally outperformed the repetitive strategy in identifying credible referent variables.
• The iterative strategy showed substantial superiority in conditions with high proportions of noninvariance or large latent variable differences.
• The iterative approach is suggested as a reliable and practical method for referent variable selection.

Conclusions:

• The iterative search strategy offers improved accuracy for referent variable selection in MIMIC-interaction models.
• This method is particularly beneficial when dealing with significant levels of measurement noninvariance.
• An illustrative example demonstrates the practical applicability of the iterative strategy.