Multilevel Meta-Analysis of Individual Participant Data of Single-Case Experimental Designs: One-Stage versus...

Area of Science:

• Psychometrics
• Biostatistics
• Research Methodology

Background:

• Multilevel meta-analysis is crucial for synthesizing findings from single-case experimental designs (SCED).
• Individual participant data (IPD) meta-analysis offers a powerful approach for analyzing SCED data.
• Two primary methods exist for IPD meta-analysis: one-stage and two-stage approaches.

Purpose of the Study:

• To explore and compare the application of one-stage and two-stage IPD meta-analysis methods for SCED.
• To evaluate the performance of these methods across SCEDs of varying complexity.
• To provide insights into the practical implementation of these techniques in statistical software.

Main Methods:

• A simulation study was conducted to compare the one-stage and two-stage IPD meta-analysis approaches.
• Multilevel models were estimated using both raw data (one-stage) and aggregated effect sizes (two-stage).
• The study examined convergence rates, estimation speed, and bias of fixed effects across different model complexities.

Main Results:

• The two-stage IPD approach demonstrated superior convergence rates for more complex single-case designs.
• Model estimation speed did not consistently favor either the one-stage or two-stage method.
• Both one-stage and two-stage IPD meta-analysis yielded unbiased point estimates for fixed effects across all simulated models.

Conclusions:

• Both one-stage and two-stage IPD meta-analysis are viable for SCED, providing unbiased fixed-effect estimates.
• The two-stage approach is recommended for complex SCED models due to better convergence properties.
• The findings support the methodological rigor of IPD meta-analysis for single-case research and discuss R implementation.