metaplus: an R package for the analysis of robust meta-analysis and meta-regression

Ken J. Beath

doi:10.32614/rj-2016-001

metaplus: an R package for the analysis of robust meta-analysis and meta-regression

Ken J. Beath

Research output: Contribution to journal › Article › peer-review

12 Citations (Scopus)

172 Downloads (Pure)

Abstract

The metaplus package is described with examples of its use for fitting meta-analysis and meta-regression. For either meta-analysis or meta-regression it is possible to fit one of three models: standard normal random effect, t-distribution random effect or mixture of normal random effects. The latter two models allow for robustness by allowing for a random effect distribution with heavier tails than the normal distribution, and for both robust models the presence of outliers may be tested using the parametric bootstrap. For the mixture of normal random effects model the outlier studies may be identified through their posterior probability of membership in the outlier component of the mixture. Plots allow the results of the different models to be compared. The package is demonstrated on three examples: a meta-analysis with no outliers, a meta-analysis with an outlier and a meta-regression with an outlier.

Original language	English
Pages (from-to)	5-16
Number of pages	12
Journal	The R journal
Volume	8
Issue number	1
DOIs	https://doi.org/10.32614/rj-2016-001
Publication status	Published - 2016

Bibliographical note

Version archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.

Access to Document

10.32614/rj-2016-001Licence: CC BY

Publisher version (open access)Final published version, 351 KBLicence: CC BY

Cite this

@article{9ebb5509a8944cf4bfc9a0dc3e83b126,

title = "metaplus: an R package for the analysis of robust meta-analysis and meta-regression",

abstract = "The metaplus package is described with examples of its use for fitting meta-analysis and meta-regression. For either meta-analysis or meta-regression it is possible to fit one of three models: standard normal random effect, t-distribution random effect or mixture of normal random effects. The latter two models allow for robustness by allowing for a random effect distribution with heavier tails than the normal distribution, and for both robust models the presence of outliers may be tested using the parametric bootstrap. For the mixture of normal random effects model the outlier studies may be identified through their posterior probability of membership in the outlier component of the mixture. Plots allow the results of the different models to be compared. The package is demonstrated on three examples: a meta-analysis with no outliers, a meta-analysis with an outlier and a meta-regression with an outlier.",

author = "Beath, {Ken J.}",

note = "Version archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.",

year = "2016",

doi = "10.32614/rj-2016-001",

language = "English",

volume = "8",

pages = "5--16",

journal = "The R journal",

issn = "2073-4859",

publisher = "Technische Universit{\"a}t Wien",

number = "1",

}

TY - JOUR

T1 - metaplus

T2 - an R package for the analysis of robust meta-analysis and meta-regression

AU - Beath, Ken J.

N1 - Version archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.

PY - 2016

Y1 - 2016

N2 - The metaplus package is described with examples of its use for fitting meta-analysis and meta-regression. For either meta-analysis or meta-regression it is possible to fit one of three models: standard normal random effect, t-distribution random effect or mixture of normal random effects. The latter two models allow for robustness by allowing for a random effect distribution with heavier tails than the normal distribution, and for both robust models the presence of outliers may be tested using the parametric bootstrap. For the mixture of normal random effects model the outlier studies may be identified through their posterior probability of membership in the outlier component of the mixture. Plots allow the results of the different models to be compared. The package is demonstrated on three examples: a meta-analysis with no outliers, a meta-analysis with an outlier and a meta-regression with an outlier.

AB - The metaplus package is described with examples of its use for fitting meta-analysis and meta-regression. For either meta-analysis or meta-regression it is possible to fit one of three models: standard normal random effect, t-distribution random effect or mixture of normal random effects. The latter two models allow for robustness by allowing for a random effect distribution with heavier tails than the normal distribution, and for both robust models the presence of outliers may be tested using the parametric bootstrap. For the mixture of normal random effects model the outlier studies may be identified through their posterior probability of membership in the outlier component of the mixture. Plots allow the results of the different models to be compared. The package is demonstrated on three examples: a meta-analysis with no outliers, a meta-analysis with an outlier and a meta-regression with an outlier.

UR - http://www.scopus.com/inward/record.url?scp=85010289375&partnerID=8YFLogxK

U2 - 10.32614/rj-2016-001

DO - 10.32614/rj-2016-001

M3 - Article

SN - 2073-4859

VL - 8

SP - 5

EP - 16

JO - The R journal

JF - The R journal

IS - 1

ER -

metaplus: an R package for the analysis of robust meta-analysis and meta-regression

Abstract

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this