Meta-analysis with Robust Variance Estimation: Expanding the Range of Working Models.

Area of Science:

• Prevention Science
• Biostatistics
• Psychology

Background:

• Large meta-analyses are prevalent in prevention science, frequently involving dependent effect size estimates.
• Current Robust Variance Estimation (RVE) methods offer meta-regression but are limited to single dependence structures.
• Existing RVE working models do not fully capture the complexity of real-world data structures.

Purpose of the Study:

• To introduce an expanded range of working models for Robust Variance Estimation (RVE).
• To present accompanying estimation methods for improved meta-regression analysis.
• To enhance the ability to capture complex data structures in meta-analyses.

Main Methods:

• Leveraging tools from multilevel and multivariate meta-analysis.
• Developing and describing new working models for dependence structures.
• Implementing methods using R packages 'metafor' and 'clubSandwich'.
• Conducting a simulation study to evaluate method performance.

Main Results:

• The proposed methods offer potential benefits for capturing diverse data structures.
• Expanded working models can improve the efficiency of meta-regression estimates.
• The study illustrates application in a meta-analysis of adolescent alcohol interventions.

Conclusions:

• Expanded RVE working models provide a more flexible and potentially efficient approach to meta-analysis.
• These methods enhance the analysis of dependent effect sizes in prevention science.
• Implementation is feasible using existing statistical software.