A robust and readily implementable method for the meta-analysis of response ratios with and without missing standard...

Area of Science:

• Ecology
• Meta-analysis
• Statistical methods

Background:

• The log response ratio (lnRR) is a common effect size in ecological meta-analyses.
• Missing standard deviations (SDs) hinder the accurate estimation of lnRR sampling variance.
• Existing methods for handling missing SDs can be imprecise.

Approach:

• Propose a new method using a weighted average coefficient of variation (CV) from studies with reported SDs.
• Estimate sampling variances for all lnRR effect sizes using this pooled CV, even when SDs are missing.
• Compare the performance of the new method against conventional approaches using simulated data.

Key Points:

• The proposed method effectively estimates sampling variances for lnRR even with missing standard deviations.
• Using a pooled CV from available studies provides more precise variance estimates than individual study-specific CVs.
• This approach minimizes bias and outperforms conventional methods, even those using complete data.

Conclusions:

• The new method for handling missing standard deviations in lnRR meta-analyses is broadly applicable.
• It offers a robust and more precise way to calculate effect sizes in ecological research.
• Implementation is straightforward for meta-analyses dealing with missing data.