[RsR] lmRob and lmrob

Dear Suresh,

I recommend using lmrob of the robustbase package.

lmRob of the robust package used to have an advantage over lmrob if
there were categorical predictors, but as of robustbase version 0.9,
this is no longer the case.

Also note that lmrob has some extensions that aim to improve
performance for small data sets (setting="KS2011"). Details and
references are given in the help files of lmrob and lmrob.control.

Best regards,

Manuel

Hi,

I am a new, naive user who needs to use one of lmrob (package robustbase)
or lmRob (package robust) to do many robust regressions to small datasets
(N~50 points). Is there any recommendation as to which one I should use or
would find easier to use (better default options, more documentation etc)
? Or are they more or less similar ?

I saw the Robust taskview, but the explicit comparison of the 2 packages
there seems quite dated, so I thought I would ask here.

Thank you very much for your work on the packages and for your help,

Suresh

_______________________________________________
R-SIG-Robust at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-robust

Manuel Koller <koller at stat.math.ethz.ch>
Seminar f?r Statistik, HG G 18, R?mistrasse 101
ETH Z?rich ?8092 Z?rich ?SWITZERLAND
phone: +41 44 632-4673 fax: ...-1228
http://stat.ethz.ch/people/kollerma/

Hi,

I am a new, naive user who needs to use one of lmrob (package robustbase)
or lmRob (package robust) to do many robust regressions to small datasets
(N~50 points). Is there any recommendation as to which one I should use or
would find easier to use (better default options, more documentation etc)
? Or are they more or less similar ?

I saw the Robust taskview, but the explicit comparison of the 2 packages
there seems quite dated, so I thought I would ask here.

Thank you very much for your work on the packages and for your help,

Suresh

Dear Manuel,

Also note that lmrob has some extensions that aim to improve
performance for small data sets (setting="KS2011"). Details and
references are given in the help files of lmrob and lmrob.control.

Thank you. I am looking at your paper right now.

I am guessing that it may be cheap and useful to increase max.it from the  
default 50 to something like 1000 (like in the lmrob.control help page).  
Is there another variable (or small number of variables) that I can  
similarly increase if I am trying to use the function in a semi-automatic  
manner on a large number of datasets ?

Very best, Suresh

You may also want to look at k.max. For the simulations in the package
vignette, lmrob_simulation.pdf, I used max.it = 500 and k.max = 2000.

Best regards,

Manuel

Dear Manuel,

Also note that lmrob has some extensions that aim to improve
performance for small data sets (setting="KS2011"). Details and
references are given in the help files of lmrob and lmrob.control.

Thank you. I am looking at your paper right now.

I am guessing that it may be cheap and useful to increase max.it from the
default 50 to something like 1000 (like in the lmrob.control help page). Is
there another variable (or small number of variables) that I can similarly
increase if I am trying to use the function in a semi-automatic manner on a
large number of datasets ?

Very best, Suresh

_______________________________________________
R-SIG-Robust at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-robust

Manuel Koller <koller at stat.math.ethz.ch>
Seminar f?r Statistik, HG G 18, R?mistrasse 101
ETH Z?rich ?8092 Z?rich ?SWITZERLAND
phone: +41 44 632-4673 fax: ...-1228
http://stat.ethz.ch/people/kollerma/

Dear Manuel and list-members,

While trying to come up to speed on this topic, I discovered this earlier  
post on this list from Olivier Renaud about a bias in the R-squared output  
 from lmRob and a function that provides a corrected R-squared value:

https://stat.ethz.ch/pipermail/r-sig-robust/2010/000290.html

Does this issue also affect lmrob ? And would that fix be also valid with  
lmrob ? (I have also contacted Olivier directly).

Thanks again,

Suresh