RE: [NMusers] How to solve it? ERROR: PARAMETER ESTIMATE IS NEAR ITS BOUNDARY

From: Mark Sale - Next Level Solutions <mark_at_nextlevelsolns.com>
Date: Wed, 18 Jul 2007 10:13:50 -0700

Jian,
 I think this is the same problem addressed by Tom Ludden recently, see his response below.  The easiest way to fix it is to use MetrumRGs NMQUAL, which fixes all known bugs in NONMEM (as well as many other useful qualification features).  Here is the web site:
http://metruminstitute.org/downloads/nmqual/

Here is Toms response if you want to do it yourself:
Mark

Dear Dr. Meineke,

You are correct, if a parameter is too different from the initial
estimate this message will be issued. Also, if a between subject
variance is close to 0 or a between subject covariance yields a
correlation close to 0, -1 or 1, then this message may be issued.  With
this knowledge, it is usually easy to identify the offending parameter
estimate(s).  This is a new "feature" that Stuart Beal added to NONMEM
VI.  Some users have had positive comments !
 while others what to
 suppress
it.  To suppress this message for most situations change the statement

DATA INCR/1E-02/   found in NONMEM routines (nm directory) named
 CN.for,
CN1.for and DECN1.for

to DATA INCR/0/

Delete all *.obj and *.lib files in the nm directory and recompile by
rerunning CDSETUP6 with the original arguments, which can usually be
found in SETUP.TXT, from the NONMEM home directory.

The Introduction to NONMEM VI does alert the user that there are new
error messages but does not provide the details.   This response will
 be
copied to nmusers so that users are better informed.

Thank you for your message.

Tom Ludden

 

Mark Sale MD
Next Level Solutions, LLC
www.NextLevelSolns.com
Received on Wed Jul 18 2007 - 13:13:50 EDT