`panel11pt7.fn` <-
function(ni=3000,nb=2000,nthin=1,nchain=3){
# R script containing WinBUGS specification of CJS model with individual 
# heterogeneity in detection and survival probabilities. Model is fitted 
# to the Dipper data. Results should be consistent with Table 11.6 in the 
# text. The MCMC algorithm should be run for 100k iterations or more due 
# to slow mixing. Mixing can potentially be improved by a number of reformulations
# of the model, but the one provided here is the most direct.
# From Royle (2008; Biometrics 64:364-370).


if(!exists("dipper.data")){
 cat("data file:  dipper.data   not found",fill=TRUE)
 return(NULL)
}

library("R2WinBUGS")
logit<-function(x){
log(x/(1-x))
}

y<-dipper.data$y
first<-dipper.data$first
nind<-dipper.data$nind
nyear<-dipper.data$nyear
Zst<-dipper.data$Zst

data <- list ("y","first","nind","nyear")

cat("
model {
for(j in 1:(nyear-2)){
yeffp[j]~dunif(0,1)
yeffp2[j]<-log(yeffp[j]/(1-yeffp[j]))
yeffphi[j]~dunif(0,1)
yeffphi2[j]<-log(yeffphi[j]/(1-yeffphi[j]))
}
yeffp[nyear-1]~dunif(0,1)
yeffp2[nyear-1]<-log(yeffp[nyear-1]/(1-yeffp[nyear-1]))
yeffphi2[nyear-1]<-0

sigma.phi~dunif(0,10)
sigma.p~dunif(0,10)
tauphi<-1/(sigma.phi*sigma.phi)
taup<-1/(sigma.p*sigma.p)


for(i in 1:nind){
eta[i]~dnorm(0,taup)  
delta[i]~dnorm(0,tauphi)
for(t in 1:(nyear-1)){
logit(p[i,t])<- yeffp2[t] + eta[i]
logit(phi[i,t])<- yeffphi2[t] + delta[i]
}
}

for(i in 1:nind){
z[i,first[i]]~dbern(1)
for(j in (first[i]+1):nyear){
muz[i,j]<-phi[i,j-1]*z[i,j-1]
z[i,j]~dbern(muz[i,j])
muy[i,j]<-p[i,j-1]*z[i,j]
y[i,j]~dbern(muy[i,j])

}
}
}
",file="cjsmodel.txt")

pst<-logit(c(.67,.87,.88,.88,.990,.73))
phist<-logit(c(.73,.45,.48,.62,.6,.73))
inits <- function(){
  list (
yeffp=runif(6,.5,.7),
yeffphi=runif(5,.5,.7),
sigma.p=runif(1,0,5), sigma.phi=runif(1,0,5),
z=Zst
)
}

parameters <- c("sigma.p","sigma.phi","yeffphi","yeffp")


out <- bugs (data, inits, parameters, "cjsmodel.txt", n.thin=nthin,n.chains=nchain, n.burnin=nb,n.iter=ni,debug=TRUE)


return(out)





}