Skip to content

[RsR] New maintainer? Package 'robust' (original S+ "Library 'robust') orphaned on CRAN

6 messages · Martin Maechler, Olivier Renaud, Valentin Todorov +2 more

Original
Martin Maechler
# 
> Dear Martin,
    > I was thinking of this since Kjell
    > abandoned the package, but I was not sure to what extent would this be
    > necessary - most of the functionality of 'robust' we already moved to
    > 'robustbase' and/or 'rrcov', therefore it suggests/imports these packages.
    > Let us see in the next days what still remains there and is valuable and
    > decide what to do.

    > Best regards,
    > Valentin

Thank you, Valentin.

I had similar thoughts a long time ago, I think prompted
indepedently by  Kjell  and Doug Martin ...

I vaguely remember I found that there have been things in
'Robust' which looked valuable and notably good to have for
comparison reasons,
*but* that were a lot work to "tear out" of the robust package
(including the necessary underlying Fortran code, but *not*
 taking all the Fortran code that would *not* be needed ...).

In the mean time the CRAN team has archived the package
'robust', and  I must adapt the 'Robustness' CRAN task view
considerably (and remove the 'Suggests: robust'  from my
robustbase package.

I'd be grateful for help and suggestions in still trying
'salvage' valuable resources out of 'robust'.

BTW:  Why has  rrcovNA  be archived?

Best,
Martin


    > On Fri, 28 Feb 2020, 11:03 am Martin Maechler, <maechler at stat.math.ethz.ch>
> wrote:
>> Dear R Robustniks,
    >> 
    >> as I am the maintainer of package 'robustbase' and am submitting
    >> a new version of it to CRAN,
    >> it came to my knowledge that package 'robust' (= the original
    >> S-plus "Library 'robust', then released to CRAN, maintained by
    >> Kjell Konis for time) has been orphaned on CRAN relatively
    >> recently.
    >> As I have 'robust' among the 'Suggests: ' entries of "my"
    >> package 'robustbase' I now get a  NOTE   about suggesting an
    >> orphaned package.
    >> I've been formally a co-author of that package (togeth
    >> 
    >> I personally would think it to be a loss for the "robustnik
    >> community" if this package would not remain easily available
    >> (notably on Windows, also the Mac where most people cannot
    >> install package from the source but are used to install the
    >> binary built package from CRAN).
    >> 
    >> Are some you willing and capable to become the maintainer of
    >> 'robust' (with the perspective to do this for a couple of years)?
    >> 
    >> Best,
    >> Martin
Olivier Renaud
# 
Dear Martin,

First of all, thank you to you, to Kjell and to all contributors for 
these two great packages. If I can add my two cents, mostly as a teacher 
to non statistically inclined students. It is confusing for the basic 
user to have robustbase and robust, to have lmrob and lmRob, covrob, 
covRob, etc.

Even if the 'Robustness' CRAN task view gives some explanations, for 
somebody who did not study robustness, it is still not clear which one 
to chose. So whatever the decision on these packages, it should be 
clearer (in the task view and in the help files) what a basic user 
should do.

I do not know enough to advise whether to merge them, or to include only 
a few functions or to let robust RIP, but here are two points:

- Concerning lmrob and lmRob I have found differences, especially in the 
initial estimator that may make a difference especially with a lot of 
factors in the x.? Liked to that the aovRob was useful

- I liked the possibility to compare 2 outputs with? fit.models (I 
believe I add to add fmclass.add.class( "lmfm", "lmrob") to make it work 
with lmrob)

Olivier
On 05/03/2020 16:20, Martin Maechler wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Valentin Todorov
    

    

      
      #
    

  


  

      
Dear Martin,

The most interesting functionality in 'robust', that I see, is the
possibility to compare  visually two or more different models.
And maybe some people are used to and prefer the interface of 'robust'. Or
some data sets.

I rebuilt it and can take over the maintenance for the next "couple of
years" as you said.

And no, rrcovNA is not archived, see
https://cran.r-project.org/web/packages/rrcovNA/index.html

Best regards,
Valentin


On Thu, 5 Mar 2020 at 17:20, Martin Maechler <maechler at stat.math.ethz.ch>
wrote:

            

      
      
      
    

        

  
  


  


      

    

    

      
      

        
      

    

    

      
      

        


  

        

      Emily Klein
    

    

      
      #
    

  


  

      
Good morning all,

A quick note just to mention that I am watching this conversation with
interest - in large part because I am in the process of using it for
analysis of ecological data. Previous papers have used lmRob(), but of
course we only have lmrob() currently, so that's what we've applied. If
both are available, I would try both and compare results, and would be very
interested in any thoughts on how they differ. I had not previously found
any documentation comparing the methods or package - could be very useful
to have that written up somewhere, but as others have said, this is a
thankless job, and I am sure you are all very busy.

Happy Friday,

Emily

        
On Thu, Mar 5, 2020 at 2:28 PM <matias at stat.ubc.ca> wrote:

        

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    
2 days later

    

      
      

        


  

        

      Martin Maechler
    

    

      
      #
    

  


  

          

      
      
      
    

        
> Dear Martin, The most interesting functionality in
    > 'robust', that I see, is the possibility to compare
    > visually two or more different models.  And maybe some
    > people are used to and prefer the interface of
    > 'robust'. Or some data sets.

    > I rebuilt it and can take over the maintenance for the
    > next "couple of years" as you said.

To close this thread:

In the mean time, Valentin Todorov, has already become the new
maintainer of 'robust'.

With many thanks to him -- and to everyone else who mentioned
what they found important in package 'robust' and not easily
available in other places:

Best,
Martin 


    > And no, rrcovNA is not archived, see
    > https://cran.r-project.org/web/packages/rrcovNA/index.html

    > Best regards, Valentin


    > On Thu, 5 Mar 2020 at 17:20, Martin Maechler

        
> <maechler at stat.math.ethz.ch> wrote:

        
>> >>>>> Valentin Todorov >>>>> on Fri, 28 Feb 2020 22:31:21
    >> +0100 writes:
    >> 
    >> > Dear Martin, > I was thinking of this since Kjell >
    >> abandoned the package, but I was not sure to what extent
    >> would this be > necessary - most of the functionality of
    >> 'robust' we already moved to > 'robustbase' and/or
    >> 'rrcov', therefore it suggests/imports these packages.  >
    >> Let us see in the next days what still remains there and
    >> is valuable and > decide what to do.
    >> 
    >> > Best regards, > Valentin
    >> 
    >> Thank you, Valentin.
    >> 
    >> I had similar thoughts a long time ago, I think prompted
    >> indepedently by Kjell and Doug Martin ...
    >> 
    >> I vaguely remember I found that there have been things in
    >> 'Robust' which looked valuable and notably good to have
    >> for comparison reasons, *but* that were a lot work to
    >> "tear out" of the robust package (including the necessary
    >> underlying Fortran code, but *not* taking all the Fortran
    >> code that would *not* be needed ...).
    >> 
    >> In the mean time the CRAN team has archived the package
    >> 'robust', and I must adapt the 'Robustness' CRAN task
    >> view considerably (and remove the 'Suggests: robust' from
    >> my robustbase package.
    >> 
    >> I'd be grateful for help and suggestions in still trying
    >> 'salvage' valuable resources out of 'robust'.
    >> 
    >> BTW: Why has rrcovNA be archived?
    >> 
    >> Best, Martin
    >> 
    >> 
    >> > On Fri, 28 Feb 2020, 11:03 am Martin Maechler, <

        
>> maechler at stat.math.ethz.ch> > wrote:

        
>> 
    >> >> Dear R Robustniks,
    >> >>
    >> >> as I am the maintainer of package 'robustbase' and am
    >> submitting >> a new version of it to CRAN, >> it came to
    >> my knowledge that package 'robust' (= the original >>
    >> S-plus "Library 'robust', then released to CRAN,
    >> maintained by >> Kjell Konis for time) has been orphaned
    >> on CRAN relatively >> recently.  >> As I have 'robust'
    >> among the 'Suggests: ' entries of "my" >> package
    >> 'robustbase' I now get a NOTE about suggesting an >>
    >> orphaned package.  >> I've been formally a co-author of
    >> that package (togeth
    >> >>
    >> >> I personally would think it to be a loss for the
    >> "robustnik >> community" if this package would not remain
    >> easily available >> (notably on Windows, also the Mac
    >> where most people cannot >> install package from the
    >> source but are used to install the >> binary built
    >> package from CRAN).
    >> >>
    >> >> Are some you willing and capable to become the
    >> maintainer of >> 'robust' (with the perspective to do
    >> this for a couple of years)?
    >> >>
    >> >> Best, >> Martin
    >>