Multistep estimators of the between-study covariance matrix under the multivariate random-effects model for...

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Estimating between-study variance is crucial for random-effects meta-analysis.
• Existing multivariate estimators are limited extensions of univariate methods.

Purpose of the Study:

• To develop novel multivariate estimators for the between-study covariance matrix.
• To extend the generalized method of moments to the multivariate setting.

Main Methods:

• Extension of the univariate generalized method of moments.
• Derivation of multivariate multistep estimators.
• Investigating the limit of multistep estimators as steps approach infinity.

Main Results:

• The proposed methodology is a viable alternative to existing methods.
• Multistep estimators perform well with heterogeneous data but not homogeneous data.
• The new estimator is semi-parametric and computationally feasible in high dimensions.

Conclusions:

• The proposed multivariate multistep estimators are a fully viable alternative for meta-analysis.
• These methods are well-suited for sensitivity analyses and heterogeneous data.
• Applicable to multivariate random-effects meta-regression.