Dear All,
I would like to reparameterize the model in a logistic Normal hierarchical
model. Specifically, I would like to place a prior on the logarithm of
precision parameters. The data consist of vectors of binary data and there is
a correlation structure within each vector.
The model is the following:(same as posting on Dec. 30, 2002)
Y_ji = 1 if species 'j' is present on island 'i',
Y_ji = 0 otherwise. i=1,...,25; j=1,2
Y_ji ~ Bernoulli(p_ji) (conditionally independent)
logit( p_ji) = beta0 + beta1*x_ji + e_ji
where (in BUGS notation) e[j,1:2] ~ dmnorm(mean0, R[, ]) j=1,...,25.
The covariate was centred a priori. I used hierarchical centering also(see
seeds example). For the precision parameters in R, I tried a variation of two
themes:
1. I placed a prior on the logarithm of each parameter in R, I specified a
non-informative prior in the following fashion:
R[1, 1] <- exp(log.Omega[1,1])
log.Omega[1,1] ~ dnorm(0,0.001), etc.
Note that I initialized log.Omega[1,1], etc.
With this formulation, BUGS was able to update up to about 500 iterations.
>From the Trap window, I think what is happening is that the matrix R fails to
be positive definite as some point as this is not checked at each iteration if
a non-Wishart prior is specified.
2. I tried a Wishart prior but could not get it to compile. The message was
'unable to choose updater for node log.Omega[1,1]'. Part of the syntax is
given below:
R[1, 1] <- exp(log.Omega[1,1])
R[2, 2] <- exp(log.Omega[2,2])
R[1, 2] <- exp(log.Omega[1,2])
R[2, 1] <- R[1,2]
log.Omega[1:2, 1:2] ~ dwish(Omega[,], 2)
Thank you for reading this long message and your help.
Melody Ghahramani
-------------------------------------------------------------------
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
|