R help - Apr 2022 - What are the pros and cons of the various R functions and methods for conducting least median of squares regression analysis?

What are the pros and cons of the various R functions and methods for
conducting least median of squares regression analysis?

I know about these:

lqs, wth method = "lms" and lmsreg, which as I understan dit are
equivalent

Mentions:
https://www.rdocumentation.org/packages/MASS/versions/7.3-56/topics/lqs

https://stat.ethz.ch/pipermail/r-help/2006-October/115681.html

https://stat.ethz.ch/pipermail/r-help/2007-March/126564.html
-----

ltsReg

https://www.rdocumentation.org/packages/robustbase/versions/0.1-2/topics/ltsReg
-----
nl.lmsNM
https://rdrr.io/cran/nlr/man/nl.lmsNM.html

No opinion (or expertise), but I think you may have missed the most
important one, Roger Koenker's package, quantreg:

https://CRAN.R-project.org/package=quantreg

Cheers,
Bert


Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )

On Sat, Apr 16, 2022 at 11:45 AM Kelly Thompson <kt1572757 at gmail.com>
wrote:
>
> What are the pros and cons of the various R functions and methods for
> conducting least median of squares regression analysis?
>
> I know about these:
>
> lqs, wth method = "lms" and lmsreg, which as I understan dit are
equivalent
>
> Mentions:
> https://www.rdocumentation.org/packages/MASS/versions/7.3-56/topics/lqs
>
> https://stat.ethz.ch/pipermail/r-help/2006-October/115681.html
>
> https://stat.ethz.ch/pipermail/r-help/2007-March/126564.html
> -----
>
> ltsReg
>
>
https://www.rdocumentation.org/packages/robustbase/versions/0.1-2/topics/ltsReg
> -----
> nl.lmsNM
> https://rdrr.io/cran/nlr/man/nl.lmsNM.html
>
> ______________________________________________
> R-help at 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.

Kelly,
I am not going to answer your question directly as I have no experience with
all?those packages and others.
Your question is very broad as you have not told us what your criteria and needs
are.
For some purposes, the package to use may be one that has been around, seems?to
continue to be supported, has had many kinks and errors straightened out and?so
on.
For other purposes, a package often comes with lots of other "related"
functionality?and often? uses names that hide other functions you may use. If a
package?also meets your other needs, or takes your data in the current format
rather than?making you rearrange first or after it returns, or other such
considerations, that may?be a valid set of criteria.
Do you care about various costs and overhead? You may not if your data is small
and?there may be a huge difference if the function does too much that you do not
need?or an inefficient algorithm. If your need is one-time, fine. If it will be
used regularly,?you may want to use some testing while measuring use of time and
memory and?anything else you care about which can include parallelization. In
many cases,?you can view the source code and sometimes even copy some and modify
for?your own use, especially if the function takes dozens of arguments you never
expect to?use and your needs are far simpler.
And, realistically, if you work with others, it may work better to use
whatever?they use, all else being close to equal.
I will say this is not the best place to ask such questions as the focus is on
base R.?Some here prefer not to have questions even about some fairly commonly
used?packages. But an abstract question about a plethora of packages, many not
mentioned,?may get more answers like mine rather than useful ones.
Good luck.


-----Original Message-----
From: Kelly Thompson <kt1572757 at gmail.com>
To: r-help at r-project.org
Sent: Sat, Apr 16, 2022 2:45 pm
Subject: [R] What are the pros and cons of the various R functions and methods
for conducting least median of squares regression analysis?

What are the pros and cons of the various R functions and methods for
conducting least median of squares regression analysis?

I know about these:

lqs, wth method = "lms" and lmsreg, which as I understan dit are
equivalent

Mentions:
https://www.rdocumentation.org/packages/MASS/versions/7.3-56/topics/lqs

https://stat.ethz.ch/pipermail/r-help/2006-October/115681.html

https://stat.ethz.ch/pipermail/r-help/2007-March/126564.html
-----

ltsReg

https://www.rdocumentation.org/packages/robustbase/versions/0.1-2/topics/ltsReg
-----
nl.lmsNM
https://rdrr.io/cran/nlr/man/nl.lmsNM.html

______________________________________________
R-help at 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]]

On Sat, 16 Apr 2022, Kelly Thompson writes:
> What are the pros and cons of the various R functions and methods for
> conducting least median of squares regression analysis?
>
> I know about these:
>
> lqs, wth method = "lms" and lmsreg, which as I understan dit are
equivalent
>
> Mentions:
> https://www.rdocumentation.org/packages/MASS/versions/7.3-56/topics/lqs
>
> https://stat.ethz.ch/pipermail/r-help/2006-October/115681.html
>
> https://stat.ethz.ch/pipermail/r-help/2007-March/126564.html
> -----
>
> ltsReg
>
>
https://www.rdocumentation.org/packages/robustbase/versions/0.1-2/topics/ltsReg
> -----
> nl.lmsNM
> https://rdrr.io/cran/nlr/man/nl.lmsNM.html
>
There is also a special mailing list for robust statistics with R:
https://stat.ethz.ch/mailman/listinfo/R-sig-Robust

As others have already suggested: you'd probably get
better answers if you ask more specific questions,
e.g. why/for what application you want to use LMS.


-- 
Enrico Schumann
Lucerne, Switzerland
http://enricoschumann.net

Thanks to all responded for your insights and advice, and for the
information about the mailing list for robust statistics with R. I
will post to that list.

I feel that I am in the early stage of learning about and "exploring"
least median of squares regression analysis and other robust
regression methods such as least absolute deviation regression, and
iteratively reweighted least squares regression, and I feel I do not
yet know enough to ask more specific questions or have the knowledge
and understanding to know what else I "should" be asking.

I am interested in robust regression methods useful for datasets with
50,000 or more observations, that might have "outliers" and/or
"influential observations", where the regression model might have one
to perhaps ten independent variables.

Here are questions I have. I'm open to suggestions of other questions
to ask, and to learning from others:
- What R functions offer least median of squares regression analysis?
- Which of these functions if any continue to be maintained?
- Are there any clear pros and cons of each package?

Thanks again.

On Sat, Apr 16, 2022 at 2:45 PM Kelly Thompson <kt1572757 at gmail.com>
wrote:
>
> What are the pros and cons of the various R functions and methods for
> conducting least median of squares regression analysis?
>
> I know about these:
>
> lqs, wth method = "lms" and lmsreg, which as I understan dit are
equivalent
>
> Mentions:
> https://www.rdocumentation.org/packages/MASS/versions/7.3-56/topics/lqs
>
> https://stat.ethz.ch/pipermail/r-help/2006-October/115681.html
>
> https://stat.ethz.ch/pipermail/r-help/2007-March/126564.html
> -----
>
> ltsReg
>
>
https://www.rdocumentation.org/packages/robustbase/versions/0.1-2/topics/ltsReg
> -----
> nl.lmsNM
> https://rdrr.io/cran/nlr/man/nl.lmsNM.html