Multiple Group Analysis in Multilevel Data Across Within-Level Groups: A Comparison of Multilevel Factor Mixture...

Area of Science:

• Multilevel modeling
• Structural equation modeling
• Statistical analysis

Background:

• Evaluating multiple group analysis methods is crucial for multilevel data.
• Existing methods require careful consideration of within-level comparison groups.
• Multilevel factor mixture models (ML FMM) and multilevel multiple-indicators multiple-causes (ML MIMIC) offer potential solutions.

Purpose of the Study:

• To evaluate the performance of ML FMM and ML MIMIC for multiple group analysis with within-level comparison groups in multilevel data.
• To assess the methods' ability to test factorial invariance (weak and strong) and latent group mean differences.
• To provide guidelines for selecting appropriate methods based on study conditions.

Main Methods:

• Two Monte Carlo simulation studies were conducted.
• Study 1 used a multilevel one-factor confirmatory factor analysis (CFA) model.
• Study 2 employed a multilevel two-factor CFA model, fitting alternative complex models.

Main Results:

• Both ML FMM and ML MIMIC demonstrated reasonable performance in multilevel multiple group analysis.
• ML MIMIC showed a slight advantage in smaller sample sizes for the one-factor model.
• For complex models, ML FMM was recommended due to the computational intensity of ML MIMIC's weak invariance testing.

Conclusions:

• The choice between ML FMM and ML MIMIC depends on the specific research model and sample size.
• Information criteria for establishing factorial invariance require careful application based on sample size.
• Guidelines are provided to aid researchers in selecting and applying these multilevel analysis methods.