Adjusting for Measurement Noninvariance with Alignment in Growth Modeling

Area of Science:

• Psychometrics
• Quantitative Psychology
• Statistical Modeling

Background:

• Longitudinal measurement invariance is crucial for valid growth modeling.
• Measurement noninvariance over time can bias growth parameter estimates.
• Existing methods often require identifying noninvariant items a priori.

Purpose of the Study:

• To extend the alignment-within-confirmatory factor analysis (AwC) technique for growth models.
• To evaluate AwC's performance in adjusting for longitudinal measurement bias.
• To compare AwC with the partial invariance modeling method.

Main Methods:

• Monte Carlo simulation study.
• Application of the alignment-within-confirmatory factor analysis (AwC) technique.
• Comparison of AwC with partial invariance modeling.

Main Results:

• AwC effectively reduces biases in growth parameter estimates.
• AwC demonstrates good control of Type I error rates, particularly with large sample sizes (N >= 1,000).
• AwC outperforms partial invariance when all items are noninvariant, but bias occurs when over 25% of parameters are noninvariant.

Conclusions:

• AwC is a viable alternative to partial invariance for growth modeling when invariance is uncertain.
• AwC offers an advantage by not requiring prior identification of noninvariant items.
• The study demonstrates AwC's utility with an example and introduces effect size indices for assessing invariance.