Hello, I am new to the BUGS mailing list but have been a WinBUGS user for several years now, intermittently.
I am estimating some fairly complex Multi-Nomial-Logit imputation models and am trying to diagnose some unusual problems I have not encountered before now.
1) A model I am running gives me a TRAP error message ONLY when I try to sample a specific node in the model. If I do not sample this node the model runs without error. Why would a model that suffers numerically only give TRAP errors when I sample interim nodes? I'm happy to provide the code if necessary. Also, the node I'm sampling is done so merely to diagnose the model, the final model parameters are sampled without problem but do not quite make sense. It seems 'problematic' for the software to 'crash' only if you decide to 'look at it'. Also, in perhaps 30 other models in WinBUGS I've never run into this problem.
2) The imputation model I'm trying to estimate at this stage imputes observations of a partially observed variable, say X[]. Currently 1/4 of the observations in X[] are observed, the rest are NAs. The NA's are currently imputed using quite informative Normal priors.
I am then wanting to use imputed and observed X[] in a subsequent model in the program. If I sample X[] I obtain only the imputed values (logically). The output of this second model seems to be consistent with only stochastic X[]s, although I can't be sure.
Is there special coding I need to use to assign both observed and imputed X's to a node in a subsequent model? Instructions associated with the 'cut' function (in program user manual) seem to imply that both stochastic and deterministic data are passed through, for example:
X.cut <- cut(X)
would assign only the stochastic data to X.cut. This seems to imply that w/o the 'cut' function both stochastic and likelihood data would be retained.
Thanks in advance for any and all help.
Professor, Department of Civil & Environmental Engineering
Arizona State University
PO Box 875306 Tempe, AZ 85287-5306
Phone: (480) 965-2220
Fax: (480) 965-0557
Express Packages:
Arizona State University
Ira A. Fulton School of Engineering
Department of Civil and Environmental Engineering
Room ECG 252 Tempe, AZ 85287-5306
-------------------------------------------------------------------
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
|