[S] How to use ltsreg (Summary)

Arman Maghbouleh (arman@csli.stanford.edu)
Fri, 4 Sep 1998 17:06:09 -0700 (PDT)


Hello, earlier I asked a question about the use of ltsreg(), least
trimmed squares robust regression. I am grateful to Brian Ripley and
Ar Menio for informative replies, summarized here.

Q. What is the relation of the $stock component returned in the
result of ltsreg() to the quan observations whose residuals were
minimized?

A. The short answer is that the two are not the same. $stock is a list
of the observations used to create the fit. The quan observations used
in minimizing the objective are the ones associated with the smallest
sized residuals ($objective == (sum(sort(($residuals)^2)[1:quan]))/quan)

Q. Is there any way to guide ltsreg() to weight some observations as
more likely to be outliers?

A. There is an open implementation of ltsreg and of S-estimation in Marazzi's
ROBETH library which one can potentially modify.

Q. Any other references?

A.

P. J. Rousseeuw and A. M. Leroy (1987) Robust Regres-
sion and Outlier Detection. Wiley.

A. Marazzi (1993) Algorithms, Routines and S Functions
for Robust Statistics. Wadsworth & Brooks/Cole.

P. Rousseeuw and M. Hubert (1997) Recent developments
in PROGRESS. In L1-Statistical Procedures and Related
Topics ed Y. Dodge, IMS Lecture Notes volume 31, pp.
201-214.

I am also looking forward to replacements for lmsreg/ltsreg/cov.mve
under development for Venables & Ripley 3rd edition.

With appreciations to the list,
Arman Maghbouleh.
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news