---------- Forwarded Message ---------- Date: 05 February 2003 14:47 +0000 From: Nicos Middleton <[log in to unmask]> To: [log in to unmask] Subject: Spatial smoothing using winBUGS Dear list Some time age I posted a question to the list but unfortunately I haven't recieved any answers. I understand my question is not of a modelling nature but rather a puzzling thing I encountered while applying the Poisson-CAR model to locally smooth area levels of mortality in Britain (around 10000 areas) e.g. Observed deaths[i] ~ dpois(mu[i]) log(mu[i]) <- log(Expected deaths[i]) + const + S[i] S[1:N] ~ car.normal(adj[], weights[], neigh[], vstar.inv) Since the CAR variance is a conditional variance, I approximate the marginal variance by calculating an empirical variance of the spatial effects S[i]'s: sdemp<-sd(S[]) varemp<- pow(sdemp, 2) However, while the conditional variance converges fine, the empirical variance does not. There is an obvious increasing trend either in all or in some chains producing a highly skewed posterior. As a consequence, on the contary of what is expected (conditional variance is approximately empirical variance * average number of neighbours), the empirical variance is greater than the conditional variance. Anyone has used such a model and has possibly observed something similar? Help will be much appreciated. Thanks, Nicos Middleton ------------------- Nicos Middleton Department of Social Medicine University of Bristol Canynge Hall Whiteladies Road Bristol BS8 2PR ---------- End Forwarded Message ---------- ---------------------- N Mitletton, Social Medicine [log in to unmask] ------------------------------------------------------------------- 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