Dear Mark
Strange behavior indeed.
Your model is overparameterized indeed (you have information to estimate
theta = n*p but you can not disantangle n and p).
This is particularly visible if you have a look on the correlation
between n and p (very high and negative)
If you add the constraint I(,10000) to p~in your OpenBUGS program
(p~dpois(lambda)I(,10000)), your will find some results which are really
simlar to WinBUGS.
I can not explain why but this is a problem that I already encountered
with Poisson distributionss when moving from WinBUGS to OpenBUGS.
I hope it helps
Etienne
Mark Strong a écrit :
> Dear BUGS users
>
> OpenBUGS 3.0.3 and WinBUGS 1.4.3 give very different results given the
> following simple model (both run on same 32 bit XP machine).
>
> WinBUGS looks correct to me. Is there a known problem with OpenBUGS?
>
> best wishes
>
> Mark Strong
>
> model
> {
> n~dnorm(mu,tau)
> p~dpois(lambda)
>
> theta<-n+p
>
> data~dnorm(theta,prec)
> }
>
> #data
> list(mu=0,tau=0.00001,lambda=5,data=5,prec=1)
>
> #inits
> list(n=0,p=5)
>
> WinBUGS results
>
> node mean sd MC error 2.5% median 97.5%
> start sample
> n -0.07729 2.396 0.08106 -5.023 0.01429 4.306
> 1000 10001
> p 5.062 2.184 0.07424 1.193 4.922 9.641
> 1000 10001
> theta 4.985 1.004 0.009822 3.037 4.984 6.964
> 1000 10001
>
> OpenBUGS results
>
> mean sd MC_error val2.5pc median val97.5pc
> start sample
> n -8.785 3.891 0.2937 -16.99 -8.633 -1.673
> 1000 10001
> p 13.78 3.749 0.291 7.0 14.0 22.0 1000 10001
> theta 4.991 0.9737 0.01066 3.164 4.982 6.946
> 1000 10001
>
>
--
------------------------------------------------------------
Etienne RIVOT
Pôle Halieutique - AGROCAMPUS OUEST Centre de Rennes
UMR INRA-Agrocampus Ecologie et Santé des Ecosystèmes
Laboratoire Ecologie Halieutique
Tel. +33. (0)2.23.48.59.34 (ou 55.36)
Fax. +33. (0)2.23.48.55.35
URL : http://www.agrocampus-ouest.fr/halieutique/
E-mail : [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
|