From: Wang, Yaning WangYA@cder.fda.gov
Subject: [NMusers] FW: [BUGS] SERIOUS WARNING: Potential error in interpreting
Date: Fri, June 4, 2004 11:39 am
Dear NONMEM users:
Sorry for the cross-maillist posting. But this may affect some PKBugs users.
Yaning Wang, PhD
Pharmacometrician
OCPB, FDA
Quoted reply history
-----Original Message-----
From: David Spiegelhalter [mailto:david.spiegelhalter@mrc-bsu.cam.ac.uk]
Sent: Friday, June 04, 2004 11:30 AM
To: BUGS@JISCMAIL.AC.UK
Subject: [BUGS] SERIOUS WARNING: Potential error in interpreting
**************************************************************
SERIOUS WARNING to all WinBUGS users: potential error in interpreting
WinBUGS model statements
**************************************************************
A recent example posted to this list showed up the following bug, as
identified by Andrew Millard.
There is a bug in the WinBUGS syntax interpreter that may in some
circumstances lead to incorrect interpretation of the following syntax:
... + (b - c) ... may be interpreted as - b + c
... - (b + c) ... may be interpreted as b - c
... - (b - c) ... may be interpreted as - b - c
Users should check their models for the existence of these structures:
essentially a bracket including a sum or difference of random quantities
with a negative sign either inside or outside the bracket. It is
perhaps most likely to occur in complex non-linear regression models.
If suspicious, try rewriting the expression in an equivalent form but
without the bracket, then running a few iterations to check it was being
evaluated appropriately.
This is clearly very serious, and we are trying to get a patch sorted
out as quickly as possible. In the meantime users are advised to check
any models they use for the occurence of this particular structure.
We do apologise for this bug, and hope that it has not led to any
inappropriate analyses. Please contact bugs@mrc-bsu.cam.ac.uk with any
queries.
David Spiegelhalter and the WinBUGS team
-------------------------------------------------------------------
This list is for discussion of modelling issues and the BUGS software.
For help with crashes and error messages, first mail bugs@mrc-bsu.cam.ac.uk
To mail the BUGS list, mail to bugs@jiscmail.ac.uk
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 jiscmail@jiscmail.ac.uk
If this fails, mail bugs-request@jiscmail.ac.uk, NOT the whole list
_______________________________________________________