Dear All,

 

I try to estimate a ZIP model, that is a mixture Poisson-binomial, specifying the following model:

 

# ZIP

 

model {

  for (i in 1 : N) {

      muSp[i,1]<-0; muSp[i,2]<-mu

      lambda[i]<-muSp[i,Index[i]]

      O[i]  ~ dpois(lambda[i])

      Index[i] ~ dcat(theta[])

      log(lambda[i]) <- log(E[i]) + alpha + h[i]

      h[i] ~ dnorm(0, tau.h)     

     }

     mu ~ dexp(0.0001); p ~ dbeta(1,1)

     theta[1]<-p; theta[2]<-1-p

}

  # Other priors:

  alpha  ~ dflat() 

  tau.h  ~ dgamma(0.5, 0.0005)      

  sigma.h <- sqrt(1 / tau.h)                      

}

 

I have, however, at least one problem. When I try to ‘compile’ the model, the WinBUGS’ message is ‘Node lambda is defined more than once’.

 

Of course, but my question is, how I could define a equation for the mean of the Poisson part of the model (lambda)?

 

Thanks for any comments or ideas

 

Marc

 

 

Prof. Marc Saez, PhD, CStat

Research Group on Statistics,

Applied Economics and Health (GRECS)

University of Girona

Campus de Montilivi, 17071 Girona

Tel 34-972418736, Fax 34-972418032

e-mail: [log in to unmask]             [log in to unmask]

 

http://www.udg.edu/fcee/economia/english/english.htm