`panel6pt3.fn` <-
function(ni=31000,nb=1000,nthin=5,nc=2,nz=60){
# This R script fits a closed population model with an individual
# covariate by data augmentation in WinBUGS. Uses the microtus data 
# with individual covariate "body mass"

library("R2WinBUGS")

y<-microtus.data[,1:5]
mass<-microtus.data[,6]
nind<-nrow(y)
J<-ncol(y)
y<-apply(y,1,sum)

# standardize covariate -- always do this.
mass<- (mass-mean(mass))/sqrt(var(mass))

# data augmentation -- pad encounter vector with zeros and covariate 
# with missing values
y<-c(y,rep(0,nz))
mass<-c(mass,rep(NA,nz))

sink("winbugsmodel.txt")
cat("model {

psi~dunif(0,1)
mu~dnorm(0,.001)
tau~dgamma(.001,.001)
sigma<-sqrt(1/tau)
a0~dnorm(0,.1)
a1~dnorm(0,.1)

for(i in 1:(nind+nz)){
 mass[i]~dnorm(mu,tau) ###I(-12,12)
 z[i]~dbern(psi)
#ff[i]<-z[i]+1
logit(p[i])<- a0+a1*mass[i]
muy[i]<-z[i]*p[i]
y[i]~dbin(muy[i],J)
}
N<-sum(z[1:(nind+nz)])
}
",fill=TRUE)
sink()



data <- list ("y","nind","nz","mass","J")
inits <- function(){
  list (
  mu=rnorm(1),
  tau=runif(1,0,5),
  a0=rnorm(1),
  a1=rnorm(1),
  psi=runif(1),z=rbinom(length(mass),1,.5))
}

parameters <- c("N","a0","a1","mu","sigma","psi")


out <- bugs (data, inits, parameters, "winbugsmodel.txt", n.thin=nthin,n.chains=nc, n.burnin=nb,n.iter=ni,debug=FALSE)







out




}