A Computationally More Efficient and More Accurate Stepwise Approach for Correcting for Sampling Error and...

Area of Science:

• Social Sciences
• Quantitative Psychology
• Organizational Research

Background:

• Multilevel structural equation modeling (ML-SEM) is widely used in social sciences for unbiased estimation.
• Maximum Likelihood (ML) is the default estimation method in Mplus for ML-SEM.
• ML can produce degenerate Level 2 (L2) covariance matrices, leading to inaccurate L2 regression coefficients with small sample sizes.

Purpose of the Study:

• To present a novel approach for stabilizing L2 covariance matrices in ML-SEM.
• To improve the accuracy of L2 regression coefficient estimates.
• To offer a more reliable method for researchers using ML-SEM with limited sample sizes.

Main Methods:

• A simulation study was conducted to compare the proposed stabilization approach with standard ML estimation.
• The proposed method focuses on stabilizing the L2 covariance matrices.
• An empirical example from organizational research is used for illustration.

Main Results:

• The proposed approach for stabilizing L2 covariance matrices yields more accurate estimates of L2 regression coefficients compared to standard ML.
• This improved accuracy is particularly evident when dealing with small sample sizes at Level 2.
• The simulation study provides evidence for the effectiveness of the stabilization technique.

Conclusions:

• The proposed method offers a valuable alternative for enhancing the precision of multilevel structural equation modeling.
• Researchers can achieve more reliable estimates in their multilevel analyses by employing this covariance matrix stabilization technique.
• This approach addresses a critical limitation of standard ML estimation in ML-SEM, particularly in data-limited scenarios.