var 
  x[N],y[N],t[N],     # pre-drug, post-drug and total PVC count
  state[N],state1[N], # binary indicator of whether patient is cured
  theta,              # probability of cure (prob of state = 1)
  p, beta,            # p = binomial probability = beta/(1+beta)
  P[2],               # `pick' variable used to select appropriate
                      # value for the binomial probability depending
                      # on whether state1 = 1 or 2 (not cured or cured)
  alpha, delta;       # p and theta transformed to logit scale for normality
model {
# MODEL
   for (i in 1:N) {
      y[i] ~ dbin(P[state1[i]], t[i]);
      state[i] ~ dbern(theta);
      state1[i] <- state[i]+1;   # state[i] takes values 0 or 1, so need to
                                 # add 1 to get values for use as index on P
   }
   P[1] <- p; P[2] <- 0;
   logit(p) <- alpha; alpha ~ dnorm(0,1.0E-4); 
   beta <- exp(alpha);  # beta measures change in rate of PVCs after treatment
   logit(theta) <- delta; delta ~ dnorm(0,1.0E-4)
}