Hello Andrew,
Thanks for this answer. In retrospect I should've seen this but I was thrown by the fact that it works in WinBUGS. I also got a reply from Kate Cowles who essentially says the same thing. So thanks to both for getting back to me.
Best,
Neil
--
Neil Walker, MSc
Statistician
The Food and Environment Research Agency
Woodchester Park,
Tinkley Lane,
Nympsfield,
Stonehouse,
GL10 3UJ
Email: [log in to unmask]
Web: www.defra.gov.uk/fera
Tel: 01453 861400
Please consider the environment before printing this e-mail
-----Original Message-----
From: andrew lawson [mailto:[log in to unmask]]
Sent: 24 November 2011 14:58
To: Neil Walker
Cc: [log in to unmask] ope this helps
Subject: Re: [BUGS] initial values for intrinsic CAR effects
Hello Neil
There is a condition on Openbugs that inits for CAR models (and possibly
other priors as well) have to sum to zero.
The way to ensure that is to specify your own inits (simplest way would
be to give 0 value to all items).
Win BUGS doesnt require this BTW
best
Andrew Lawson
On 11/23/2011 12:13 PM, Neil Walker wrote:
> Hello BUGS,
> I am currently trialling and refining a model in OpenBUGS in which one of the priors is an intrinsic CAR distribution. If I randomly generate initial values for the CAR effects (I simulated from a Normal with mean 0 and a small variance for all CAR effects), the analysis stalls and I get a message to the effect "initial value for car[1] not feasible". I can get round this by setting all initial values for CAR effects to zero but it would be preferable if I could vary these in the usual way. The other thing to add is that I can successfully run the same model in WinBUGS with randomly generated CAR effects.
> Am I missing something very obvious?
> Any help appreciated.
> regards,
> Neil
>
> --
> Neil Walker, MSc
> Statistician
>
> The Food and Environment Research Agency
> Woodchester Park,
> Tinkley Lane,
> Nympsfield,
> Stonehouse,
> GL10 3UJ
>
> Email: [log in to unmask]
> Web: www.defra.gov.uk/fera
> Tel: 01453 861400
>
>
> Please consider the environment before printing this e-mail
>
>
>
> The information contained in this message may include privileged, proprietary or confidential information. Please treat it with the same respect that you would expect for your own information. If you have received it in error, we apologise and ask that you contact the sender immediately and erase it from your computer. Thank you for your co-operation.
>
> The original of this email was scanned for viruses by the Government Secure Intranet virus scanning service. On leaving the GSi this email was certified virus free.
>
> -------------------------------------------------------------------
> 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
The information contained in this message may include privileged, proprietary or confidential information. Please treat it with the same respect that you would expect for your own information. If you have received it in error, we apologise and ask that you contact the sender immediately and erase it from your computer. Thank you for your co-operation.
The original of this email was scanned for viruses by the Government Secure Intranet virus scanning service. On leaving the GSi this email was certified virus free.
-------------------------------------------------------------------
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
|