R devel - Jul 2008 - Editing the "..." argument

Dear all,

I'd like tweaking the ... arguments that one user can pass in my 
function for fitting a model. More precisely, my objective function is 
(really) problematic to optimize using the "optim" function. 
Consequently, I'd like to add in the "control" argument of the
latter
function a "ndeps = rep(something, #par)" and/or "parscale =
something"
if the user has not specified it already.

Do you know a way to deal with this point?
In advance, thanks.

Mathieu

-- 
Institute of Mathematics
Ecole Polytechnique F?d?rale de Lausanne
STAT-IMA-FSB-EPFL, Station 8
CH-1015 Lausanne   Switzerland
http://stat.epfl.ch/
Tel: + 41 (0)21 693 7907

On 05/07/2008 1:55 PM, Mathieu Ribatet wrote:
>     Dear all,
> 
> I'd like tweaking the ... arguments that one user can pass in my 
> function for fitting a model. More precisely, my objective function is 
> (really) problematic to optimize using the "optim" function. 
> Consequently, I'd like to add in the "control" argument of
the latter
> function a "ndeps = rep(something, #par)" and/or "parscale =
something"
> if the user has not specified it already.
> 
> Do you know a way to deal with this point?
> In advance, thanks.
It's generally not necessary to edit the ... args, but one way to do it 
is like this:

   dots <- list(...)
   # do some editing
   # instead of optim(a, b, ...), now you do
   do.call(optim, c(list(a, b), dots))

In the particular example you gave, I wouldn't do this; I'd give control
as an arg to your function, and just edit that.  For example,

myoptim <- function(control = list(), ...) {
   if (is.null(control$ndeps)) control$ndeps <- rep(something, par)
   if (is.null(control$parscale)) control$parscale <- something
   optim(control=control, ...)
}

Duncan Murdoch

Mathieu Ribatet wrote:
>    Dear all,
> 
> I'd like tweaking the ... arguments that one user can pass in my 
> function for fitting a model. More precisely, my objective function is 
> (really) problematic to optimize using the "optim" function. 
> Consequently, I'd like to add in the "control" argument of
the latter
> function a "ndeps = rep(something, #par)" and/or "parscale =
something"
> if the user has not specified it already.
> 
> Do you know a way to deal with this point?
> In advance, thanks.
> 
> Mathieu
Package lmom, recently added to CRAN, contains a function 'pelp' that
does exactly what you describe.  Relevant parts of its code:

   # Get the user-supplied arguments for the optimization functions
   dotargs <- list(...)

   # If user didn't supply a "control" argument, create one
   if (is.null(dotargs$control)) dotargs<-c(dotargs,list(control=list()))

   # If user didn't provide "ndeps" or "abstol" elements
of the "control"
   # argument, set them to our own defaults
   if (is.null(dotargs$control$ndeps))
     dotargs$control <- c(dotargs$control,list(ndeps=rep(acc,nshape)))
   if (method!="L-BFGS-B" && is.null(dotargs$control$abstol))
     dotargs$control <- c(dotargs$control,abstol=acc^2)

   # Call the optimization function, optim()
   opt <- do.call(optim,c(list(par=start[shape.pos]),fn=critfn,
     method=method,dotargs))


J. R. M. Hosking

The 'modifyList' function on package Utils allows to do that in a very
compact way:

do.call("optim", modifyList(list(<your default values>),
list(...)))

Regards,

Enrique

Mathieu Ribatet wrote:
> Dear all,
> 
> I'd like tweaking the ... arguments that one user can pass in my 
> function for fitting a model. More precisely, my objective function is
> (really) problematic to optimize using the "optim" function. 
> Consequently, I'd like to add in the "control" argument of
the latter
> function a "ndeps = rep(something, #par)" and/or "parscale =
something"
> if the user has not specified it already.
> 
> Do you know a way to deal with this point?
> In advance, thanks.
> 
> Mathieu