[ADMB Users] NaN objective function value in report

[Chris Gast]

Hello,

As part of a series of simulations, I've been using the reporting feature of
ADMB to output the objective function value for model selection purposes.
 I'm beginning to actually use this value as my simulations progress, and
I've noticed that some simulations of some of my models produce an objective
function value of "nan" in the report document.  The accompanying parameter
output file correctly lists the numeric objective function value as that
which is last seen on the output buffer (where I print it, for each
iteration).  This occurs rarely, and for those simulations for which this
does not occur, the two objective function values match.

Has anyone else encountered this, or can anyone explain this?  Does anyone
know of a way to output the "correct" objective function value to the report
file, so I don't have to parse it out from the parameter estimate file?

Thanks very much,

Chris Gast




-----------------------------
Chris Gast
cmgast at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.admb-project.org/pipermail/users/attachments/20100715/a8ca8981/attachment.html>