Use of the estimated intraclass correlation for correcting differences in effect size by level

Area of Science:

• Meta-analysis methodology
• Psychometrics
• Biostatistics

Background:

• Standardized mean differences (d-effect sizes) in meta-analyses can be computed at individual or cluster levels.
• Cluster-level d-effect sizes may be inflated and require correction using the intraclass correlation (ICC).
• ICC values are often not reported in primary studies, hindering accurate meta-analysis.

Purpose of the Study:

• To propose a novel method for estimating ICC values from reported standard deviations within a meta-analysis.
• To provide a solution for situations where default ICC values are unavailable.
• To enable accurate correction of cluster-level d-effect sizes.

Main Methods:

• Developed a method to estimate ICC from standard deviations within meta-analytic data.
• Conducted simulations to evaluate the accuracy and precision of the estimated ICC.
• Examined the influence of factors like the number of studies on the estimation.

Main Results:

• The proposed method accurately and precisely estimates ICC values.
• The estimated ICC can be reliably used to correct inflated cluster-level d-effect sizes.
• Simulation results support the validity of the estimation technique.

Conclusions:

• The proposed method offers a practical approach to estimate ICC when not directly reported.
• Accurate ICC estimation facilitates the correct calculation of d-effect sizes in meta-analyses.
• This contributes to more robust and reliable findings in intervention and group comparison studies.