Ex 6.10 N-of-1 analysis - very robust to alternative priors on tau.
Just uniform reported below.

model;
{
for(k in 1:23){                        # j indexes individual
	dummy[k]<-id[k]+dose[k]  #  just to mention all data     
	for(j in 1:n[k]){                  # k indexes observation  
		   y[k,j]       ~ dnorm(theta[k],inv.sigma2[k]) 
		   }
    theta[k]           ~ dnorm(mu.theta, inv.tau.theta2)  
# next 2 lines used in dose-adjustment model
#	 theta[k] <- theta.adj[k]+
#	               beta*(log.dose[k]-mean(log.dose[1:23]))	      #    theta.adj[k] ~ dnorm(mu.theta, inv.tau.theta2)
	 P[k] <- step(theta[k])  # prob(effect > 0)		 
   inv.sigma2[k]  ~ dlnorm(mu.sigma,inv.tau.sigma2)#  log-normal precisions  
	   }
	mu.theta  ~ dunif(-10,10)      # uniform on soveral mean
	beta  ~ dunif(-10,10)    # used in dose-adjustment model
	mu.sigma  ~ dunif(-10,10)     # uniform on mean log-precision
	inv.tau.sigma2 <- 1/(tau.sigma*tau.sigma)
	tau.sigma  ~ dunif(0,10)       # uniform on sd of log-precisions
	P.pop <- step(mu.theta)
	P.pop.diff <- step(mu.theta-0.5)
	mean.var<- exp(mu.sigma+tau.sigma*tau.sigma/2)
 
# j again indexes individual

for(j in 1:23){	
  log.dose[j] <- log(dose[j])         # log of dose
  s[j] <-  sd(y[j,1:n[j]])            # empirical sd 
	prec2[j]  <- n[j]/(s[j]*s[j])  # empirical 1/se^2
	}

s02 <- 1/mean(prec2[1:23])   # mean for shrinkage
sigma2 <- s02 * mean(n[1:23])  # mean for ICC

# Prior: Uniform(0,50) on tau
tau.theta           ~ dunif(0,50)
tau.theta2    <- tau.theta * tau.theta
inv.tau.theta2 <- 1/tau.theta2
}