Robust Meta-Analysis and Meta-Regression

Performs meta-analysis and meta-regression using standard and robust methods with confidence intervals based on the profile likelihood. Robust methods are based on alternative distributions for the random effect, either the t-distribution (Lee and Thompson, 2008 or Baker and Jackson, 2008 ) or mixtures of normals (Beath, 2014 ).


Changes in metaplus version 0.7-11

o fixed bug in parsing which caused crash in normal profiling

Changes in metaplus version 0.7-10

o fixed warning from deprecated use of arrays

Changes in metaplus version 0.7-9

o changed options for integration of t distribution models so hopefully Gauss-Hermite is never needed

o modified Hessian calculations to avoid possible bug

o added warning message about study second parameter

o updated vignette and added code

Changes in metaplus version 0.7-8

o updated citation to reflect publication in R Journal

o updated vignette

Changes in metaplus version 0.7-7

o improved fitting for mixture models

o changed ordering of study labels for outlier probability plot to match ordering of forest plot

o added digits to parameters passed through from plot to summary lines

Changes in metaplus version 0.7-6

o added profile improvement introduced in 0.7-5 for normal models

o added option to perform integration in the t-distribution models using adaptive Gauss-Hermite quadrature to improve accuracy with studies with small standard errors

o correctly name CDP studies

o documentation enhancements

Changes in metaplus version 0.7-5

o added data parameter to metaplus to indicate where data is, and changed all examples and documentation to use this

o added a print method for metaplus objects which prints summary

o documentation changes

o renamed outlier.probs to outlierProbs and test.outliers to testOutliers to align with standard R naming conventions

o added further improvements to handing of tau^2 close to zero

o if new maxima is found during profiling then update solution (in practice this usually means that the likelihood is almost flat and very messy)

o add check for multimodal profile likelhood and give warning

o return profile so that this can be further examined

Changes in metaplus version 0.7-4

o added even more imports

Changes in metaplus version 0.7-3

o fixed occasional problems when tau^2 is zero for all distributions

o if likelihood is mis-shapen and CI plot requested then give a warning rather than fail

o check for inadequate number of studies and stop, or warn if only just adequate

o add cehck for appropriate model when calculating outlier probabilities

o add additional imports in NAMESPACE to pass R CMMD check

Changes in metaplus version 0.7-2

o fixed bug with fitting t-distribution when tau^2 was zero

o update documentation

Changes in metaplus version 0.7-1

o fixed bug related to singularity of Hessian

o when better solution found when profiling then use this fitted model

Changes in metaplus version 0.7-0

o improved convergence by using Nelder-Mead if nlminb failed

o used modified bbmle to fix bugs - changes will eventually be incorporated into bbmle

Changes in metaplus version 0.6-1

o improve profiling of t-distribution models by using multiple starting values

o fix problem when vinv=0 correctly

o fix documentation errors

o include examples vignette

Changes in metaplus version 0.6-0

o initial release

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


1.0-2 by Ken Beath, 7 months ago

Browse source code at

Authors: Ken Beath [aut, cre] , Ben Bolker [aut] , R Development Core Team [aut]

Documentation:   PDF Manual  

Task views: Meta-Analysis, Robust Statistical Methods

GPL (>= 2) license

Imports bbmle, metafor, boot, methods, numDeriv, MASS, graphics, stats, fastGHQuad, lme4, Rfast, parallel, doParallel, foreach, doRNG

Suggests R.rsp

Suggested by statsExpressions.

See at CRAN