Meta-analyzing dependent correlations with correction for artifacts that multiplicatively attenuate the true...

Area of Science:

• Psychometrics
• Statistical Methodology
• Meta-Analysis

Background:

• Existing meta-analysis methods for dependent correlations can distort effect size variation.
• Samplewise-adjusted procedures improve dependent correlation meta-analysis but lack artifact correction.
• Artifact correction (e.g., unreliability) is increasingly vital in meta-analytic research.

Purpose of the Study:

• To extend samplewise-adjusted procedures for meta-analyzing dependent correlations with artifact correction.
• To evaluate the performance of new procedures under various conditions via Monte Carlo simulation.

Main Methods:

• Monte Carlo simulation was employed to assess meta-analytic procedures.
• Simulated data varied in correlation dependence, heterogeneity, sample size, and number of studies.
• Procedures were compared based on bias and confidence interval coverage for population parameters.

Main Results:

• Previous methods, including uncorrected samplewise procedures, produced biased estimates and poor confidence interval coverage.
• Bias and undercoverage worsened with larger sample sizes and more studies.
• Newly developed samplewise-adjusted procedures with artifact correction demonstrated negligible bias in estimating mean population correlation.

Conclusions:

• Accurate meta-analysis of dependent correlations necessitates artifact correction for attenuation.
• The proposed samplewise-adjusted procedures offer a robust solution for this challenge.
• Further research can explore conditions for enhancing these improved meta-analytic techniques.