Skip to content

[RsR] R-SIG-Robust Digest, Vol 20, Issue 2

3 messages · Ajay Shah, Kjell Konis, Martin Maechler

Original
Ajay Shah
# 
Jennifer had asked this question a few days ago, and I'm also curious
about the answer. (Or, did someone answer this and I missed it?)
Kjell Konis
# 
It's probably because the Robust Library has only been available for R  
for a little over a year and only available under the GPL for about 6  
months.  Also, I don't think they "recommend rlm over lmRob" but  
rather don't mention lmRob at all.

Kjell
On 17 Nov 2007, at 18:26, Ajay Shah wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Martin Maechler
    

    

      
      #
    

  


  

          

      
      
      
    

        
KK> It's probably because the Robust Library has only been
    KK> available for R for a little over a year and only
    KK> available under the GPL for about 6 months.  
Indeed!
(and quite a few of us only learned about the GPL part two weeks
 ago, at the very nice Banff/BIRS workshop on 
 "Robust Statistics and R").

    KK> available under the GPL for about 6 months.  Also, I
    KK> don't think they "recommend rlm over lmRob" but rather
    KK> don't mention lmRob at all.

nor 'lmrob()' from package "robustbase" (an early version of
which existed as roblm() in package 'roblm') which has been
available for around two years, and has always had a faster
algorithm and more modern/accurate inference parts  than  rlm(),
similar to lmRob() mentioned above. 

As Kjell has implicitely mentioned, rlm() has been around much
longer -- from 'MASS' which is part of Venables & Ripley's 'VR' 
all of which has been available as ``standard R'' since MASS / VR
has been among the "Recommended Packages" as long as these
exist.

Some of the book authors mentioned below really come from an
S-plus background rather than R, and others are not much focused
on robustness statistics; so are somewhat behind "current state
of the art"  on  ``robust statistics using R''.

Martin Maechler, ETH Zurich

        
KK> On 17 Nov 2007, at 18:26, Ajay Shah wrote:

        
>>> First, several books that discuss R and have mentioned
    >>> robust methods (Crawley, 2007; Faraway, 2005; Fox, 2002;
    >>> Jureckova & Picek, 2006) have used the rlm function in
    >>> the MASS library. Do you have any idea why these books
    >>> would recommend rlm over lmRob? I don?t find the output
    >>> to be as helpful as lmRob?s output. Well, this is just
    >>> curiosity.
    >> 
    >> Jennifer had asked this question a few days ago, and I'm
    >> also curious about the answer. (Or, did someone answer
    >> this and I missed it?)

Yes, indeed you did miss Kjell's earlier answer.

    >> Ajay Shah http://www.mayin.org/ajayshah
    >> ajayshah at mayin.org http://ajayshahblog.blogspot.com
    >> <*(:-? - wizard who doesn't know the answer.