Christophe & Michael,

I know I didn't provide much context to my question hence it might seem weird.

But here is an example: I am trying to predict my response variable,
in a longitudinal set up,

y.pred[i] <- exp(b[1] + b[2]*treat[i] + W.pred[subject[i]] + U.pred[i])

where W.pred and U.pred are predicted values. Also we are assuming
that U.pred follows a mixture of uniforms.

I would have loved to provide the whole code and include data but I am
restricted in that sense. So I merely picked out the relevant parts.

Thanks for responding.


Regards,
Raphael





On Wed, Sep 26, 2012 at 8:18 AM, Christophe Saint-Jean
<[log in to unmask]> wrote:

Hi Raphael,

   Some weird question :  Are you sure that a mixture of uniform is

identifiable (in the sense defined in the book "Finite Mixture Models") ?

Best,

Christophe.

Le 26 sept. 2012 à 12:02, Raphael Fraser a écrit :

Hi,

My goal is to sample from a mixture of uniforms but unless U[j] is observed

or u[j] is a number my code will not work. Is there an alternative? Can

anyone help?

Thanks,

Raphael

model {

       for (j in 1:M) {

                       U[j] ~ dunif(-u[j],u[j])

                       K[j] ~ dcat(p[1:N])

                       u[j] <- r[K[j]]

                       for (i in 1:N) {

                               SC[j,i] <- equals(i,K[j])

                       }

               }

                       for (i in 1:N) {

                               sumSC[i] <- sum(SC[,i])

                               cl[i] <- step(sumSC[i] - 1)

                       }

                       NC <- sum(cl[])

       # Contructive DPP

       for (k in 1:N) { r[k] ~ dgamma(1.5,tau.e) }

       for (k in 1:(N-1)) { V[k] ~ dbeta(1,1) }

       for (k in 2:N) { p[k] <- V[k]*(1-V[k-1])*p[k-1]/V[k-1] }

       V[N] <- 1

       p[1] <- V[1]

       # priors

       tau.e <- pow(2*pow(sig.e,2),-1)

       sig.e ~ dunif(0,6)

}

#### INITS

list(sig.e=3.0)

#### DATA

list(M=10,N=5)

------------------------------------------------------------------- This

list is for discussion of modelling issues and the BUGS software. For help

with crashes and error messages, first mail [log in to unmask] To mail

the BUGS list, mail to [log in to unmask] Before mailing, please check the

archive at www.jiscmail.ac.uk/lists/bugs.html Please do not mail attachments

to the list. To leave the BUGS list, send LEAVE BUGS to

[log in to unmask] If this fails, mail [log in to unmask], NOT

the whole list

--------------------------------------------

     Christophe Saint-Jean

  Lab. MIA, Dép. Informatique

    Université de La Rochelle

   http://perso.univ-lr.fr/csaintje/

--------------------------------------------