Hi,
I've been fitting generalised linear mixed models using BUGS 0.6, with
quite small simulated genetic data sets (50x 4-generational pedigrees, each
with 13 members). In quite a few cases, the MCMC process appears to
converge after a short burn-in, run for anywhere between 5000-30000 or more
iterations, then suddenly all parameters increase rapidly (from values of
about 3 to 100 in < 100 iterations) and the process crashes soon after.
Has anyone else experienced situations like this? The process looks
convergent and passes convergence tests during most of the run.
Thanks for any help or suggestions.
Cheers,
Katrina
****************************************************************************
Katrina Scurrah
PhD student
Division of Biostatistics and Genetic Epidemiology
TVW Telethon Institute for Child Health Research
PO Box 855
West Perth WA 6872
Phone: (08) 9388 2703
Fax: (08) 9388 2734
email: [log in to unmask]
http://www.ichr.uwa.edu.au
****************************************************************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|