Hello,
I experimented the Metafor and Meta packages in the scope of replacing Excel
for meta-analysis. I performed the first working example provided in Michael
Borenstein's book "Introduction to Meta-Analysis" with Excel, Metafor and Meta.
The numbers given by my spreadsheet, which I validated from Borenstein's book,
conrespond quite closely to those given by Meta, but are different from those
obtained using Metafor. For the fixed effect, I infer that the differences are
related to numerical issues, but for the random effect, the numbers are
considerably different. Unfortunately, I could not find where I made it wrong.
I would be grateful if someone would have a look at my calculations.
Here are the meta-analysis commands:
### USING METAFOR
library(metafor)
( dat<-escalc(m1i=m1i, sd1i=sd1i, n1i=n1i, m2i=m2i, sd2i=sd2i, n2i=n2i,
measure="SMD", data=metaData, append=T) ) # COMPUTE EFFECT SIZE
( res<-rma.uni(yi,vi,data=dat,method="HE", level=95) ) ### RANDOM EFFECT
( res<-rma.uni(yi,vi,data=dat,method="FE", level=95) ) ### FIXED EFFECT
### USING META
( res<-metacont(metaData[,3], metaData[,1], metaData[,2], metaData[,6],
metaData[,4], metaData[,5],
studlab=rownames(metaData),sm="SMD",
level = 0.95, level.comb = 0.95,
comb.fixed=TRUE, comb.random=TRUE,
label.e="Experimental", label.c="Control",
bylab=rownames(metaData)) )
The whole R script is temporarly available at http://bit.ly/eYesbZ
The spreadsheet is temporarly available at http://bit.ly/fAYWPo
Kind regards,
S.-?. Parent, Eng., Ph.D.
Department of Soils and Agrifood Engineering, Universit? Laval
Canada
Metafor vs Meta vs Spreadsheet: wrong numbers
2 messages · Serge-Étienne Parent, Viechtbauer Wolfgang (STAT)
1 day later
Dear Serge-?tienne, Try: res <- rma.uni(yi, vi, data=dat, method="DL") for the random-effects model when using the metafor package. You used method="HE" -- this will give you a different estimator of tau^2 then the one used in the meta package. Best, -- Wolfgang Viechtbauer Department of Psychiatry and Neuropsychology School for Mental Health and Neuroscience Maastricht University, P.O. Box 616 6200 MD Maastricht, The Netherlands Tel: +31 (43) 368-5248 Fax: +31 (43) 368-8689 Web: http://www.wvbauer.com ----Original Message---- From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of Serge-?tienne Parent Sent: Wednesday, January 12, 2011 18:18 To: r-help at r-project.org Subject: [R] Metafor vs Meta vs Spreadsheet: wrong numbers
Hello,
I experimented the Metafor and Meta packages in the scope of
replacing Excel for meta-analysis. I performed the first working
example provided in Michael Borenstein's book "Introduction to
Meta-Analysis" with Excel, Metafor and Meta. The numbers given by my
spreadsheet, which I validated from Borenstein's book, conrespond
quite closely to those given by Meta, but are different from those
obtained using Metafor. For the fixed effect, I infer that the
differences are related to numerical issues, but for the random
effect, the numbers are considerably different. Unfortunately, I
could not find where I made it wrong. I would be grateful if someone
would have a look at my calculations.
Here are the meta-analysis commands:
### USING METAFOR
library(metafor)
( dat<-escalc(m1i=m1i, sd1i=sd1i, n1i=n1i, m2i=m2i, sd2i=sd2i,
n2i=n2i, measure="SMD", data=metaData, append=T) ) # COMPUTE EFFECT
SIZE ( res<-rma.uni(yi,vi,data=dat,method="HE", level=95) ) ###
RANDOM EFFECT ( res<-rma.uni(yi,vi,data=dat,method="FE", level=95) )
### FIXED EFFECT
### USING META
( res<-metacont(metaData[,3], metaData[,1], metaData[,2],
metaData[,6], metaData[,4], metaData[,5],
studlab=rownames(metaData),sm="SMD", level = 0.95, level.comb =
0.95, comb.fixed=TRUE, comb.random=TRUE, label.e="Experimental",
label.c="Control", bylab=rownames(metaData)) )
The whole R script is temporarly available at http://bit.ly/eYesbZ
The spreadsheet is temporarly available at http://bit.ly/fAYWPo
Kind regards,
S.-?. Parent, Eng., Ph.D.
Department of Soils and Agrifood Engineering, Universit? Laval Canada