[RsR] estimators based on random samples... - should be random

Werner Stahel @t@he| @end|ng |rom @t@t@m@th@ethz@ch
Wed May 3 14:01:22 CEST 2006


I clearly advocate the use of R's random numbers throughout.

I think that an argument seed= is not needed, since the handling
of random numbers is easy without it. The user can call set.seed
before calling lmrob and, if needed, store the old seed and
reset .Random.seed to this value later.

As to the naming issue, I dislike the underline as a part of a
name, because I use ESS. I prefer names like
lmrobMM
lmrobS

Is there a need for method="fastS"? Will there be other
submethods of S-estimators? If so, I again like the main things
first:  Sfast  instead of  fastS .

Greetings
Werner 
----------------- This message was sent by ---------------------------
Werner Stahel                              http://stat.ethz.ch/~stahel
Seminar fuer Statistik                     phone  :    +41 1 632 34 30
ETH-Zentrum, LEO D8                        fax    :    +41 1 632 12 28
CH-8092 Zurich, Switzerland                meet me: Leonhardstr.27, D8




More information about the R-SIG-Robust mailing list