A Monte Carlo Simulation Study on the Influence of Unequal Group Sizes on Parameter Estimation in Multilevel Confirmatory Factor Analysis

ABSTRACT

Unequal group sizes (imbalance) and small sample sizes are common in multilevel confirmatory factor analyses (ML-CFA). This simulation study examined the influence of imbalance combined with small sample sizes on both levels on estimation performance in ML-CFA. Imbalance did not influence estimation performance given the minimum sample size requirements. Greater sample sizes on one level compensated for smaller sample sizes on the respective other level. Additionally, the degree of intraclass correlation ($ICC$) interacted with sample sizes. Based on the results of the simulation study, recommendations for practical applications are delineated. For instance, at least 100 Level-2 units with an average cluster size of four or 150 Level-2 units with an average cluster size of two are recommended given an $ICC$ of .30 or above.

KEYWORDS:

• Multilevel confirmatory factor analysis
• unequal groups
• imbalance
• Monte Carlo simulation

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Note that in contrast to the presented ML-CFA model, ML-CFA models in both studies included a factor structure on the between-level. They are cited because among the available literature they resemble the MTMM-ML-CFA model for interchangeable raters the most.

Additional information

Funding

The authors received no financial support for this work.