[R] reproducing an error in dumped frames

[Andy Jacobson]

Greetings,

        I am doing many iterations of simple optimizations using the
        L-BFGS-B method of optim() in R-1.6.2.  Very rarely (about 1
        in every 100,000 trials), a process will encounter some
        problem in optim that triggers the error message "non-finite
        value supplied by optim".  From the C code, it appears that
        this message could be emitted in several different places.

        I have tried to capture the exact conditions that cause this
        error via setting options(error) to call dump.frames().  I've
        gotten a frame dump or two now, and so I believe I have a few
        sets of guilty function arguments.  However, when I call
        optim() with these arguments, I cannot make it crash.  Am I
        wrong to have expected to be able to reproduce the problem in
        this manner?  Any suggestions would be much appreciated!

        Thanks,
                Andy

-- 
Andy Jacobson

andyj at splash.princeton.edu

Program in Atmospheric and Oceanic Sciences
Sayre Hall, Forrestal Campus
Princeton University
PO Box CN710 Princeton, NJ 08544-0710 USA

Tel: 609/258-5260  Fax: 609/258-2850