Hi, all.
I am combining proportions from two experiments in a hierarchical model:
pi = ri/Ni, with logit(pi) distributed as b[I] ~ norm(mu, tau). This is
very similar to the surgical institution ranking. When ri's are
non-zero, works fine. When one of the ri's is zero, I get an overflow
error. Any help appreciated. Is Poisson model more appropriate?
MODEL
model
{
for( i in 1 : N ) {
b[i] ~ dnorm(mu,tau)
r[i] ~ dbin(p[i],n[i])
logit(p[i]) <- b[i]
}
mu ~ dnorm(0.0,1.0E-6)
sigma <- 1 / sqrt(tau)
tau ~ dgamma(0.001,0.001)
}
Data
list(n = c(1194, 4796),
r = c(0,4),
N = 2)
Inits
list(b = c( 0.01, 0.01),
tau = 1, mu = 0)
Daniel
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|