[R] non linear 3SLS with constraints
spencer.graves at pdf.com
Thu Feb 2 18:40:29 CET 2006
RSiteSearch("3SLS") produced 27 hits for me just now. This
identified several functions in contributed packages that offer 3SLS. I
didn't look at all of them, but I didn't see any mention of constraints
in the ones I skimmed. RSiteSearch("nonlinear least squares with
constraints") returned 13, including
"http://finzi.psych.upenn.edu/R/Rhelp02a/archive/13304.html". This is a
comment from Doug Bates, who is one of the world's leading experts on
nonlinear regression. He suggests eliminating the constraints by
transformation, and provides a reference to his book on the subject
(which I highly recommend, if you are not already familiar with it).
I can't think of any sensible technique related to statistical
computing that can NOT be done in R. Moreover, the more obscure or
nonstandard the technique, the more likely it will be more easily done
in R than anything else. For me, R is the premier platform for new
statistical algorithm development. Much of what is available in other
packages is already available in R or in a contributed package.
Without this listserve, R has a pretty steep learning curve.
However, that difficulty is substantially reduced by the ready
availability of an international network of people who contribute their
time to helping others via this listserve. Some questions are answered
in minutes. A few never receive an answer. Questions that don't get
answered often seem to require specialized knowledge that relatively few
people have. For example, I don't know what percentage of those who
follow this listserve know '3SLS', but I'd guess it might be rather low.
If you are persistent, recasting a question in alternative, more
generic terms, or decribing briefly why you want to do something (not
just asking about a specific technique you've heard about), you will get
an answer. (I believe the posting guide,
"www.R-project.org/posting-guide.html", is also quite helpful.)
In particular, if you think you need 3SLS rather than some more
general nonlinear least squares or maximum likelihood, or if you don't
think transformations are appropriate, please submit another question to
this listserve explaining as clearly and succinctly as you can why you
hope this helps.
Van Campenhout Bjorn wrote:
> i am new here and wanted to know, before i start learning yet another statistical package:
> I want to estimate a system of equations that is
on linear in the parameters, using 3SLS. However,
i will probably have to constrain some of the parameters
to be between, say, zero and one. Is this possible with
R, and better, is this easy? If so, since i am an absolute
beginner, any pointers where to start would be greatly appreciated.
> [[alternative HTML version deleted]]
> R-help at stat.math.ethz.ch mailing list
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
More information about the R-help