[R-meta] Tau2 estimate problem

[Viechtbauer, Wolfgang (SP)]

Hi Mutlu,

The problem is in this line:

y1=c(rnorm(study.nr, mean = eff.size, sd = t2+s2))

Note that 'sd' is for specifying the SD. So it should be:

y1=c(rnorm(study.nr, mean = eff.size, sd = sqrt(t2+s2)))

Then it will work as expected.

Best,
Wolfgang

-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On Behalf Of Mutlu Umaroglu
Sent: Thursday, 10 October, 2019 14:29
To: r-sig-meta-analysis using r-project.org
Subject: [R-meta] Tau2 estimate problem

Dear All,

I generated random data for meta analysis with user-defined effect size 
and tau2 (eg. es=1, T^2=0.3). After generating random data, I perform a 
meta-analysis. However, the results (for tau2) are different from 
user-defined tau2. I need some technical assistance to fix code.

Regards,
Mutlu


rowMeans(es2)		#1
[1] 0.9941907 0.9941133

> rowMeans(tu2)		#0.3
[1] 0.08509195 0.08346508

----

library(metafor)
library(Sim.PLFN)

set.seed(9999)

sml.nr=1000		#1000 simulation
study.nr=10		#with 10 studies in meta analysis

eff.size=1		#Effect size for meta analysis
t2=.3			#Tau2 for meta analysis

es2<-NULL
tu2<-NULL

for (i in 1:sml.nr){

   s2=Trunc(n=study.nr, T.dist="chisq", T.dist.par=1, L=0.36, R=2.4)/4
   y1=c(rnorm(study.nr, mean = eff.size, sd = t2+s2))
   metadata<-data.frame(y1,s2)

	DLmet<-rma(y1,s2,data=metadata,method="DL")
	REMLmet<-rma(y1,s2,data=metadata,method="REML")

	EffectSize= c(DLmet$b,REMLmet$b)
         es2 = cbind(es2, EffectSize)

	taau2= c(DLmet$tau2,REMLmet$tau2)
	tu2 = cbind(tu2, taau2)
}