A Tool Box for Multivariate Meta-Analysis

Description

The package xmeta consists of a collection of functions for making inference and detecting publication bias in multivariate meta-analysis (MMA).

Details

Inference

The aim of the estimation methods is to estimate the coefficients β and the components of the between-study (co)variance matrix Ψ for multivariate random-effects meta-analysis. One major challenge in MMA is the standard inference procedures, such as the maximum likelihood or maximum restricted likelihood inference, require the within-study correlations, which are usually unavailable. Different estimators with and without the knowledge of within study correlation are implemented in the package xmeta. The estimation methods available in function mmeta are:

• Restricted maximum likelihood for MMA with continuous outcomes

• Composite likelihood method for MMA with continuous outcomes

• Method of moment for MMA with continuous outocmes

• Improved method for Riley model for MMA with continuous outcomes

• Marginal bivariate normal model for MMA with binary outcomes

• Marginal beta-binomial model for MMA with binary outcomes

• Hybrid model for disease prevalence along with sensitivity and specificity for diagnostic test accuracy

• Trivariate model for multivariate meta-analysis of diagnostic test accuracy

Publication bias

Detecting and accounting for publication bias are challenging in MMA setting. The multivariate nature is often not fully accounted for by the existing univariate methods. The score test for detecting publication bias in MMA when the within-study correlations are unknown is implemented in the function mpbt. The Galaxy method for correcting publication bias in MMA is implemented in the function galaxy.

Author(s)

Yong Chen
Chuan Hong
Maintainer: Yong Chen <ychen.upenn@gmail.com> Contributor: Haitao Chu, Yulun Liu

Results