A new approach for handling missing correlation values for meta-analytic structural equation modeling: Corboundary R...

Area of Science:

• Multivariate statistical analysis
• Meta-analysis methodology
• Data imputation techniques

Background:

• Multivariate meta-analysis is increasingly used across disciplines to examine complex variable relationships.
• Missing correlation data within correlation matrices (R) presents a significant challenge in these analyses.
• Existing methods may not adequately address the derivation of accurate prior distributions for missing correlations.

Purpose of the Study:

• To address the challenge of missing correlations in multivariate meta-analysis.
• To establish more informative and empirical prior distributions for missing correlation coefficients (r).
• To develop a method for deriving mathematical boundaries for missing correlations within a valid correlation matrix.

Main Methods:

• Discusses the mathematical/analytical derivation of boundaries for missing correlations (r).
• Ensures derived correlations satisfy conditions for a valid correlation matrix (symmetric, positive semidefinite, real values between -1 and 1).
• Demonstrates an R package for constructing empirical distributions of these correlation boundaries.

Main Results:

• Provides a method to analytically determine boundaries for missing correlation coefficients (r).
• The R package facilitates the construction of empirical distributions for these boundaries.
• The approach ensures the integrity and validity of the correlation matrix (R) during imputation.

Conclusions:

• The study offers a novel approach to handle missing correlations in multivariate meta-analysis.
• Empirical prior distributions derived from mathematical boundaries improve data imputation.
• The methodology has potential applications beyond multivariate meta-analysis for correlation matrix completion.