Quantcast

Error in JAGS, cannot monitor z

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Error in JAGS, cannot monitor z

alon.benari
Hello.
 I am running a state space model (the example from Marc kery's Bayesian population analysis using WinBUGS )

The code runs fine for the examle as shown in the book.
 However  I played around and changed the nodes monitored to the latent variable z  I get the following  error- even though it is monitored in the code:


Initializing model

  |++++++++++++++++++++++++++++++++++++++++++++++++++| 100%
Error in jags.samples(model, variable.names, n.iter, thin, type = "trace",  :
  Failed to set trace monitor for z
Node not found


The code is:


model {

# Priors and constraints
for (i in 1:nind){
   for (t in f[i]:(n.occasions-1)){
      phi[i,t] <- alpha[t]
      p[i,t] <- beta[t]
      } #t
   } #i
   
for (t in 1:n.occasions-1)
{
        alpha[t] ~ dbeta(1, 1)         # Prior for mean survival
        beta[t] ~ dbeta(1, 1)           # Prior for mean recapture
}  

                 # Prior for mean recapture

# Likelihood
for (i in 1:nind){
   # Define latent state at first capture
   z[i,f[i]] <- 1
   for (t in (f[i]+1):n.occasions){
      # State process
      z[i,t] ~ dbern(mu1[i,t])
      mu1[i,t] <- phi[i,t-1] * z[i,t-1]
      # Observation process
      y[i,t] ~ dbern(mu2[i,t])
      mu2[i,t] <- p[i,t-1] * z[i,t]
      } #t
   } #i
}
ni <- 10000
nt <- 6
nb <- 5000
nc <- 3


cjs.ran <- jags.model(file="cjs22time.jags",data=bugs.data, n.chains = nc,n.adapt=2000)
coda.cjs<-coda.samples(cjs.ran,c("z"),n.iter=5000,thin=5)


Any idea why would not trace z node?

Thank you.
Loading...