Multilevel meta-analysis of single-subject experimental designs: a simulation study.

Area of Science:

• Psychology
• Behavioral Science
• Research Methodology

Background:

• Single-subject experimental designs (SSEDs) are crucial for evaluating interventions.
• Combining data from multiple SSEDs enhances statistical power and generalizability.
• Multilevel meta-analysis offers a framework for synthesizing SSED data.

Purpose of the Study:

• To evaluate the performance of multilevel meta-analysis for SSED data.
• To compare the utility of unstandardized versus standardized regression coefficients as effect sizes.

Main Methods:

• A simulation study was conducted to assess multilevel meta-analysis performance.
• Unstandardized and standardized regression coefficients were used as effect size metrics.
• Performance was evaluated based on estimation accuracy and statistical power.

Main Results:

• Multilevel meta-analysis using unstandardized effect sizes yielded accurate effect estimates.
• Multilevel meta-analysis with standardized effect sizes performed well only with >= 20 measurement occasions.
• Sufficient power for intercept effects was achieved with homogeneous studies or large sample sizes.
• Sufficient power for slope effects required large numbers of studies and measurement occasions.

Conclusions:

• Unstandardized effect sizes are recommended for multilevel meta-analysis of SSEDs.
• Standardized effect sizes in this context require substantial longitudinal data.
• Study homogeneity and sample size are critical for detecting treatment effects in meta-analyses.