[RsR] OGK covariance estimator
Matias Salibian-Barrera
m@t|@@ @end|ng |rom @t@t@ubc@c@
Mon Dec 5 23:23:16 CET 2005
Martin Maechler wrote:
> \item I plan to add ``\emph{psi - Function}'' objects, using an S4 class,
> with instances for Huber, Hampel, Biweight, etc, at least for the
> M-estimators (of location / regression).
> Such an object should contain $\rho$, $\psi$, $w(x) := \psi(x)/x$,
> $\psi'$ (= $d/\mathrm{dx} \psi$) functions, default values for the tuning
> constants, and functionals such as $E_X[\psi(X)^2]$, $E_X[\psi'(X)]$,
> for $X\sim N(0,1)$ (these functionals are still \emph{functions} of the
> tuning parameters).
> [snip]
> and actually I've done the S4 class for "psi-function" objects
> last Friday.
>
> Now I'm interested to hear more..
I think estimating functions are "natural" candidates to be treated as
S4 objects, as Martin's blueprint shows above.
One question: I have written the "workhorses" of my code in C and the
"psi" functions are defined inside the C code (tuning constants are
taken from R). If we move towards having the "psi" functions constructed
in R (which I think is a good idea for many reasons), is there an easy
way to have the C code use them? .Call() comes to mind, but of course, I
would want to re-write the least possible number of lines of my original
C code.
Matias
PS: A website collecting the working groups minutes and other related
material to this project is already being set up and will be announced
in this list when it becomes "live" (which should happen soon).
More information about the R-SIG-Robust
mailing list