[6350] in s-news-athena
nlminb
daemon@ATHENA.MIT.EDU (D.Verotta)
Wed Feb 1 14:42:46 1995
Date: Wed, 1 Feb 95 11:17:40 PST
From: david@c255.ucsf.EDU (D.Verotta)
To: s-news@stat.wisc.edu
-> nlminb users:
->
-> my call to nlminb looks like
->
-> nlminb( start=rep(0.05,3) , objective=log.like , X=X , y=y , n=n )
->
-> and I get the error message:
->
-> Error in .Fortran("divset",: subroutine divset: Argument 2 has zero length
-> Dumped
->
-> :-(
->
-> I looked at the lnminb code. It looks like the second argument to divset
-> is beyond my control. Any takers on this on?
I had a similar problem in a different context. It looks like many
variables in nlminb can take NULL values and sometimes this screws up
calls to functions
I solved my problem (which sounded much similar, argument with zero legth)
by modifying the call to nlminb.init
temp <- nlminb.init(start, objective, gradient, hessian, scale, control, p, nderiv, liv, lv, lh, ...)
to:
temp <- nlminb.init(start, objective, gradient, hessian, scale, control, p = p, nderiv = nderiv, liv = liv, lv = lv, lh = lh, ...)
Unfortunately I now need to keep a local version of nlminb
davide
-----------------------------------------------------------------------------
Davide Verotta, Dr, PhD e-mail: davide@c255.ucsf.edu
Ast Prof Pharmacy and Biostatistics Phone: (415) 476-1556
University California at San Francisco
521 Parnassus Avenue, San Francisco, CA 94143-0446
Fax: (415) 476-1556
Fax alternative: (415) 476-1556 plus dial #11 after the machine picks up
Fax alternative (if none of the above works): (415) 476-9330
