#15 Bivariate Poisson model

[Anonymous]

I have found JAGSusing rjags does a lot of thengs that Winbugs could not but I hit a brick wall with this example
Can anyone spot my error

cat(
    "model{
    for(i in 1:n)
    {
    LH[i]<-eps[i]*AS[HT]*DS[AT]*HGA
    LA[i]<-eps[i]*AS[AT]*DS[HT]
    X[i,1] ~dpois(LH[i])
    X[i,2] ~dpois(LA[i])
    }
       for(i in 1:n)
    {
     eps[i]~dgamma(1,1)
    }


  for(Z in 1:20)
    {
 AS[Z]~dgamma(1,1)
    }


for(Z in 2:20)
{ DS[Z]~dgamma(1,1)
    }
   DS[1] <- 1
   HGA~dgamma(1,1)
    }

    ", file="BP.txt")

jags.data=NULL
jags.data$X=X
jags.data$n=n
jags.data$HT=HT
jags.data$AT=AT

jags3.m <- jags.model( file = "BP.txt", data=jags.data, 
                       inits=list("AS"=rep(1,20),"DS"=c(NA,rep(1,19)),"eps"=rep(1,n)), n.chains=3, n.adapt=500 )

I get this error message

Error in jags.model(file = "BP.txt", data = jags.data, inits = list(AS = rep(1,  : 
  RUNTIME ERROR:
Cannot insert node into LA[1]. Dimension mismatch

[Martyn Plummer]

We can't know for certain without seeing your data, but the most likely explanation is that in the expression

LA[i]<-eps[i]*AS[AT]*DS[HT]

one of AS[AT] and DS[HT] is a vector. Hence the expression evaluates to a vector and cannot be inserted into the scalar expression on the left hand side.