[RsR] Questions about interpreting lmRob output
Kjell Konis
kon|@ @end|ng |rom @t@t@@ox@@c@uk
Sun Nov 18 16:10:31 CET 2007
The Robust Library Users' Guide is in the source package for the
Robust Library. You can download it from here:
http://cran.r-project.org/src/contrib/robust_0.3-0.tar.gz
The subsampling done in lmRob is not bootstrapping. By default lmRob
will use enough subsamples to achieve a breakdown point of 0.5 with
high probability. See the help file for lmRob.control for the
particulars.
Kjell
On 15 Nov 2007, at 19:27, Jenifer Larson-Hall wrote:
> Kjell,
> Again, thanks for all your help. Could you indulge two more
> questions from me?
>
> First, I thanked you for the document you sent, but you said it was
> available to anyone in the source version of the robust library. I
> have looked around for how to get to that. I tried
> vignette("robust") but that didn't get me anything from the robust
> library. When I Googled it I got a website, http://cran.r-project.org/doc/packages/
> , that took me to a document called robust.pdf. This was, however,
> NOT the same as the document you sent me! Specifically, the document
> at that URL contains only a compilation of all of the help files, as
> far as I can see, whereas the document you sent (S-PLUS 6 Robust
> Library User's Guide Version 1.0, dated May 2002, which you noted
> was a bit old now) was more book-like and I have to say, more
> helpful than the help files. Don't get me wrong, the help files are
> helpful, but this just gave more detail than the help files ever
> did. So . . . is this S-PLUS document available somewhere else that
> I haven't figured out how to get to? I'm asking because I would like
> to be able to refer people to it. In particular, is there some
> command I am missing that I could use in R to bring up these pdf
> documents (besides the vignette command, and I do think the
> vignettes are often of the same kind of quality as the article you
> sent, more article-like or book-like, and thus more helpful to me!).
>
> Second, you didn't answer one of my questions about whether the
> resampling done in lmRob is bootstrapping. My guess, after reading
> your document (the S-PLUS) one, is that it is not the same thing,
> because the details about random resampling on p. 54 of that
> document seem to indicate that if I had 5 parameters, there would be
> 147 resamples, which is not nearly how many I am used to seeing with
> bootstrapping. So my question is what kind of resampling is this?
> Can you refer me to a source to understand better what it is if it
> is not bootstrapping? And if I changed the parameters to be 1000
> reps, would that mean I was doing bootstrapping?
>
> Thanks for your help.
> Jenifer
>
> Dr. Jenifer Larson-Hall
> Assistant Professor of Linguistics
> University of North Texas
> (940)369-8950
>
>
>
[[alternative HTML version deleted]]
More information about the R-SIG-Robust
mailing list