## Hierarchical logistic regression document.SUBSCRIPTION_OPTIONS = { "thing": "thread", "subscribed": false, "url": "subscribe", "icon": { "css": "fa fa-envelope-o" } };

Help
Anonymous
2011-02-18
2012-09-01
• Anonymous - 2011-02-18

Hi,

I am currently learning to use JAGS with rjags. I am starting by trying to
replicate some of the WinBUGS code and results from Royle and Dorazio's
Hierarchical Modeling and Inference in Ecology. The current model I am trying
to fit is a fairly simple hierarchical logistic regression. I am getting
stumped by an error message:

Error in node y

Observed node inconsistent with unobserved parents at initialization

At other times the error has occured in y or y, but these three nodes are the
first 3 non-zero nodes in the y variable. The model specification I am using
is:

```model{
for(i in 1:M){
z[i]~ dbern(psi[i])
psi[i] <- ilogit( b0+b1*elev[i]+b2*elev2[i]+b3*forest[i])
y[i]~dbin(z[i]*p,J[i])
}
p~dunif(0,1)
b0~dnorm(0,.001)
b1~dnorm(0,.001)
b2~dnorm(0,.001)
b3~dnorm(0,.001)

}
```

I have posted my r code and the data here:
https://public.me.com/pokey54

Any help in interpreting or troubleshooting this error is appreciated.

Thanks!

• Martyn Plummer - 2011-02-18

In your code you are randomly initializing the latent variable z to be 0 or 1.
But if z = 0 then y = 0. The problem arises because you've got some non-zero y
values with zero z.

To fix the code, modify your initial values for z from this

```z=rbinom(M,1,.4)
```

to this

```z=ifelse(y>0, 1, rbinom(M, 1, 0.4))
```