I am having trouble with the following model.  I have observations, x[i], that I would like to model as coming from a mixed distribution, which has a point mass at zero with probability p, and is beta(alpha, beta) otherwise ( i.e., with probability 1-p).

 

I have tried a construct like the following:

for(i in 1:N) {

x[i] ~ p*C

}

p ~ dbern(q)

q ~ dunif(0,1)

C ~ dbeta(alpha, beta)

}

 

WinBUGS will not allow this, responding that there are multiple definitions of node UA[1].  I have checked the archives, and this question has been posted, but I can't find any answers.

--
Dana Kelly
Idaho National Laboratory