Using collaboration networks to identify authorship dependence in meta-analysis results.

Area of Science:

• Bibliometrics
• Network Analysis
• Meta-Analysis

Background:

• Meta-analytic methods are crucial for synthesizing research evidence across academic disciplines.
• Sources of bias and heterogeneity in meta-analyses can arise from study authorship and researcher bias.
• Understanding research group influence is vital for accurate interpretation of meta-analytic findings.

Purpose of the Study:

• To develop an objective method for attributing study authorship to research groups within meta-analyses.
• To investigate the impact of research group origin on effect size estimates.
• To demonstrate a method for correcting bias stemming from authorship dependence in meta-analyses.

Main Methods:

• Graph cluster analysis of scientific collaboration networks to identify research groups.
• Empirical examination of how research group origin affects effect sizes in various meta-analyses.
• Application of multilevel random-effects models to account for authorship dependence.

Main Results:

• A novel method for objectively assigning studies to research groups within meta-analyses was established.
• Empirical evidence demonstrated that the research group of origin significantly impacts effect size estimates.
• Non-independence of within-group results can introduce bias if not addressed.

Conclusions:

• Graph cluster analysis provides an objective means to identify research groups in meta-analyses.
• Multilevel random-effects models effectively correct for authorship dependence, enhancing the reliability of meta-analytic results.
• This methodology improves the accuracy and reduces bias in synthesizing scientific evidence.