[RsR] Rcmdr and robust tools

Ian Fellows ||e||ow@ @end|ng |rom uc@d@edu
Tue Aug 4 18:38:44 CEST 2009


Martin,

It depends a little on the context. If you are using Deducer from the
windows Rgui, the adding and deleting items is pretty easy via the winMenu*
family of functions. If you are using JGR menus and menu items can be added
with jgr.addMenu and jgr.addMenuItem, but there is currently no way to
arbitrarily remove a menu without jumping into Java (though there is a
preference item to suppress Deducer's menus at start-up. If you are calling
Deducer from the terminal via the deducer() function which brings up a text
menu, there is no way to customize. In the future it would be a good idea to
have a unified plug-in API.


HCCM - Heteroscedasticity-Corrected Covariance Matrices
see hccm function in package car. Perhaps it is mostly Neto that uses this
terminology.

You are right that perhaps this is not what the original poster is looking
for, as it is not outlier resistant. 

What kinds of customizations do you make in RCmdr for classes that you
teach? Do you add whole dialogs, or just edit the menu options?

Ian



-----Original Message-----
From: Martin Maechler [mailto:maechler using stat.math.ethz.ch] 
Sent: Tuesday, August 04, 2009 8:45 AM
To: Fellows, Ian
Cc: 'Eva Cantoni'; 'r-sig-robust'
Subject: Re: [RsR] Rcmdr and robust tools

Hi Eva & Ian & other robustniks,

>>>>> "IF" == Ian Fellows <ifellows using ucsd.edu>
>>>>>     on Tue, 4 Aug 2009 08:26:43 -0700 writes:

    IF> Hi Eva,
    IF> I'm not sure about Rcmdr, but I just released the Deducer package to
CRAN
    IF> which uses HCCM by default with linear models. The online manual
gives some
    IF> screenshots, but I have yet to write the regression page.

Hmm, one of the acronym definition pages on the web says

>> HCCM
>> 
>> Acronym	Definition

>> HCCM		Healthcare Continuum Model
>> HCCM		Discharge Under Honorable Conditions, Convenience of the Man

But let's assume you mean heteroscedasticity correction as
provided e.g. by package 'sandwich'.

Note that using that *not* robust estimation in the sense of
most people reading this mailing list.

But now to the topic of 'Deducer' vs 'Rcmdr' (note: Subject amended: "+r") :
One reason 'Rcmdr' has been quite nice for teaching has been
that you can somewhat easily *modify* it,
by adding your own menu entries; modify or even omit existing ones.

How easy is that with 'Deducer' ?

Regards,
Martin Maechler, ETH Zurich

    IF> Manual:
    IF> http://www.deducer.org/pmwiki/pmwiki.php?n=Main.DeducerManual


    IF> Cheers,
    IF> Ian Fellows




More information about the R-SIG-Robust mailing list