Hierarchical linear model (varying intercepts model) with t distribution instead of normal

gistfile1.txt

model
{
# The model for each observational unit 
# (each row is a subject's data point)
for( j in 1:N )
{
mu[j] <- beta[1] + beta[2] * ( so[j] ) + u[subj[j]] + w[item[j]]
rt[j] ~ dnorm( mu[j], tau.e )
##change the above line to:
#rt[j] ~ dt(mu[j],tau.e, 2)
##generate predicted values: 
 rt.pred[j] ~ dnorm(mu[j],tau.e)
##change the above line to:
#rt.pred[j] ~ dt(mu[j],tau.e, 2)
 }
# Random effects for each subject
for( i in 1:I )
{
u[i] ~ dnorm(0,tau.u)
}
 # Random effects for each item
 for( k in 1:K )
 {
 w[k] ~ dnorm(0,tau.w)
 }
# Uninformative priors:
# Fixed effect intercept and slope
beta[1] ~ dnorm(0.0,1.0E-5)
beta[2] ~ dnorm(0.0,1.0E-5)
# Residual (within-person) variance
tau.e <- pow(sigma.e,-2)
sigma.e ~ dunif(0,2000)
# Between-person variation
tau.u <- pow(sigma.u,-2)
sigma.u ~ dunif(0,500) 
 # Between-item variation
 tau.w <- pow(sigma.w,-2)
 sigma.w  ~ dunif(0,500)

## track predicted values:
smallest <- min(rt.pred)
mn      <- mean(rt.pred)
largest  <- max(rt.pred)