#Bayesian random effects meta-analysis for Risk Ratio 
model
{
#Likelihood
for (i in 1:k)
{
rc[i] ~ dbin(pic[i],nc[i])
rt[i] ~ dbin(pit[i],nt[i])
mu[i] <- log(pic[i])
log(pit[i]) <- mu[i] + min(delta[i],-log(pic[i]))
delta[i] ~ dnorm(delt,precision.tau)
pic[i] ~ dunif(0,1)				#flat prior distribution for pic[i].
}
delt ~ dnorm(0,0.1)			#prior distribution of ƒÂ= ln(RR)
precision.tau <- 1/tau.squared
tau.squared <- tau*tau
tau ~ dunif(0,2)
RR<-exp(delt)
#predictive distribution
delt.new~dnorm(delt,precision.tau)	#Predicted delta

RR.new<-exp(delt.new)					#Predicted RR
#probability of delt more than 0 (RR more than 1)
pdelt0<-equals(min(delt,0),0)

#probability of delt.new more than 0 (RR.new more than 1)
pdelt.new0<-equals(min(delt.new,0),0)
}

#probability of delt.new more than 0 (RR.new less than 0.8)
temp<-log(0.8)-delt.new
pdelt.new0.8<-equals(min(temp,0),0)
}


DATA	#k is the number of studies.
list(k=8,rc=c(155,30,109,107,110,32,41,10),nc=c(381,42,184,154,139,87,69,10),rt=c(333,22,103,88,96,45,40,8),nt=c(769,42,190,148,138,99,68,10))