Dear Bugs Users
Due to (yet another!) spate of `remove' messages, the list is being moderated
again. This will result in delays before posting of messages, (e.g. outside of
office hours, UK time) and for the time being all messages will be forwarded by
me. Apologies for any frustration or inconvenience this causes.
Also, please find below a query from Alan Cobo-Lewis
Ken Rice
Bugs list manager
------------- Begin Forwarded Message -------------
I'm trying to use winbugs to fit a logistic model with cross-classified
random effects, and it's way too slow. Each of the first few iterations
takes 13-15 sec on my 850 MHz Pentium 3, so I don't dare try
1000s of them!
I have S=672 subjects with binary responses on each of I=493 items.
For inits on the subj[] and item[] residuals, I supplied the
values that I obtained when I did separate logistic regressions
on each subject and each item.
In the following model, agemo is the subject's age in months (and is
positively related to p[i,s]) and qfemale codes 1 for female and 0 for
male (females do better than males).
model
{
for (i in 1:I) {
for (s in 1:S) {
r[i,s] ~ dbern(p[i,s])
logit(p[i,s]) <- alpha0 + alpha1*agemo[s] + alpha2*qfemale[s] + subj[s] +
item[i]
}
}
for (s in 1:S) {
subj[s] ~ dnorm(0.0, tau.s)
}
for (i in 1:I) {
item[i] ~ dnorm(0.0, tau.i )
}
alpha0 ~ dnorm(0.0, 1.0E-6)
alpha1 ~ dnorm(0.0, 1.0E-6)
alpha2 ~ dnorm(0.0, 1.0E-6)
tau.s ~ dgamma(0.001, 0.001)
sigma.s <- 1/sqrt(tau.s)
tau.i ~ dgamma(0.001, 0.001)
sigma.i <- 1/sqrt(tau.i)
}
Any suggestions? Should I try reparameterizing, or do I just
have do darn much data to expect to solve this problem in a
reasonable amount of time?
alan
--
Alan B. Cobo-Lewis, Ph.D. (207) 581-3840
University of Maine fax 581-1953
5773 South Stevens Hall, Room 234B
Orono, ME 04469-5773 [log in to unmask]
http://www.umaine.edu/visualperception
------------- End Forwarded Message -------------
-------------------------------------------------------------------
To mail the BUGS list, mail to [log in to unmask]
You can search old messages at www.jiscmail.ac.uk/lists/bugs.html
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
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]
