This is the mail archive of the
gsl-discuss@sourceware.org
mailing list for the GSL project.
Re: Robust linear least squares
- From: Dirk Eddelbuettel <edd at debian dot org>
- To: Peter Teuben <teuben at astro dot umd dot edu>
- Cc: "gsl-discuss\ at sourceware dot org" <gsl-discuss at sourceware dot org>, Patrick Alken <patrick dot alken at Colorado dot EDU>
- Date: Sun, 12 May 2013 13:13:35 -0500
- Subject: Re: Robust linear least squares
- References: <518D6E3B dot 8080503 at colorado dot edu> <518FD7B4 dot 8070100 at astro dot umd dot edu>
On 12 May 2013 at 13:56, Peter Teuben wrote:
| Patrick
| I agree, this is a useful option!
|
| can you say a little more here how you define robustness. The one I
| know takes the quartiles Q1 and Q3 (where Q2 would
| be the median), then define D=Q3-Q1 and only uses points between
| Q1-1.5*D and Q3+1.5*D to define things like a robust mean and variance.
| Why 1.5 I don't know, I guess you could keep that a variable and tinker
| with it.
| For OLS you can imagine applying this in an iterative way to the Y
| values, since formally the errors in X are neglibable compared to those
| in Y. I'm saying iterative, since in theory the 2nd iteration could have
| rejected points that should have
| been part or the "core points". For non-linear fitting this could be a
| lot more tricky.
There is an entire "task view" (ie edited survey of available packages)
available for R concerning robust methods (for model fitting and more):
http://cran.r-project.org/web/views/Robust.html
So there is not just one generally accepted best option. That said, having
something is clearly better than nothing. But let's properly define the
method and delineat its scope/
Dirk
--
Dirk Eddelbuettel | edd@debian.org | http://dirk.eddelbuettel.com