[R] Question "breakdown point" --> robust statistics

Martin Maechler m@ech|er @end|ng |rom @t@t@m@th@ethz@ch
Fri Apr 20 09:08:16 CEST 2018


>>>>> Olivier Crouzet <olivier.crouzet using univ-nantes.fr>
>>>>>     on Thu, 19 Apr 2018 18:13:30 +0200 writes:

    > Hi,
    > I think he's talking about how much a statistical estimator is influenced by extreme datapoints, e.g.

    > https://en.m.wikipedia.org/wiki/Robust_statistics#Breakdown_point

    > Olivier

Definitely!  Thank you, Olivier.

I had not replied because the question looked more like homework
than something for which R-help is appropropriate.

There is the 'robustbase' package and more of a dozen other high
quality R packages about robust statistics -- in which
"breakdown point" is one crucial concept.

BTW: What econometricians tend to call "robust" is the idea
of using "sandwich" estimates (--> 'sandwich' package in R), aka
Newey-West aka HAC (Heteroscadisticity And Correlation adjusted) estimates.

Robust statistics in a proper sense considers "sensitivity" of
both estimates and their precision estimates in situations where
the distributional assumption (e.g. Gaussian errors) is only
approximately true, say 95% of the observations/errors/.. are Gaussian but the
remaining 5% can be anything (-> arbitrary extreme outliers).

There's the CRAN task view on robust statistical methods:
https://cran.r-project.org/web/views/Robust.html

Martin Maechler
ETH Zurich

    > -- 
    > Olivier Crouzet 
    > Assistant Professor
    > @LLING UMR6310 - Université de Nantes / CNRS
    > Guest Scientist
    > @UMCG - University Medical Center Groningen / RijksUniversiteit Groningen

    >> Le 19 avr. 2018 à 11:00, Keith Jewell <Keith.Jewell using campdenbri.co.uk> a écrit :
    >> 
    >>> On 15/04/2018 17:26, Marc Girondot via R-help wrote:
    >>>> Le 15/04/2018 à 17:56, alireza daneshvar a écrit :
    >>>> break-down point
    >>> Can you explain more what you plan to do and give an example of what you have tried to do until now to do a "break down point" in R. Perhaps a "break down point" is common in your field, but I have no idea about what it is !
    >>> https://en.wikipedia.org/wiki/Breakdown_(1997_film)
    >>> Sincerely
    >>> Marc
    >> Perhaps the OP means "breakpoint" in which case the 'strucchange' package might be relevant <https://cran.r-project.org/web/packages/strucchange/index.html>
    >> 
    >> ______________________________________________
    >> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
    >> https://stat.ethz.ch/mailman/listinfo/r-help
    >> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
    >> and provide commented, minimal, self-contained, reproducible code.

    > [[alternative HTML version deleted]]

    > ______________________________________________
    > R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
    > https://stat.ethz.ch/mailman/listinfo/r-help
    > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
    > and provide commented, minimal, self-contained, reproducible code.




More information about the R-help mailing list