R help - Jan 2003 - polynomial contrasts in R

In S-Plus, I can obtain polynomial contrasts for an ordered factor with
contr.poly(). The function also exists in R, however is limited to factors
where the levels are equally spaced. In S-Plus, one can obtain the contrasts
for a set of numeric values representing unequally spaced ordered factors.
Has anyone implemented this in R? I see that the S-Plus function calls
another function (poly.raw()) that calls a Fortran routine. Thanks for your
assistance.

Cheers, George

=================================================================George W.
Gilchrist                        Email #1: gwgilc at wm.edu
Department of Biology, Box 8795          Email #2: kitesci at cox.net
College of William & Mary                    Phone: (757) 221-7751
Williamsburg, VA 23187-8795                    Fax: (757) 221-6483

On 16 Jan 2003 at 17:19, George W. Gilchrist wrote:
> In S-Plus, I can obtain polynomial contrasts for an ordered factor with
> contr.poly(). The function also exists in R, however is limited to factors
> where the levels are equally spaced. In S-Plus, one can obtain the
contrasts
> for a set of numeric values representing unequally spaced ordered factors.
> Has anyone implemented this in R? I see that the S-Plus function calls
> another function (poly.raw()) that calls a Fortran routine. Thanks for your
> assistance.
> 
This can be done "by hand" easily in R. From some notes of mine:

dummy <- poly(c(1,2,3,5), 3)
   dimnames(dummy) <- list(c(1,2,3,5), c(".L", ".Q",
".C"))
    # orthogonal polinomial over the set {1,2,3,5}

   muscles$Ltreat <- C(muscles$Ltreat, dummy, 3)
   
   dummy <- poly(c(1,3,6), 2)
   dimnames(dummy) <- list(c(1,3,6), c(".L", ".Q"))
   muscles$Ntreat <- C(muscles$Ntreat, dummy, 2)

Kjetil Halvorsen


> Cheers, George
> 
> =================================================================>
George W. Gilchrist                        Email #1: gwgilc at wm.edu
> Department of Biology, Box 8795          Email #2: kitesci at cox.net
> College of William & Mary                    Phone: (757) 221-7751
> Williamsburg, VA 23187-8795                    Fax: (757) 221-6483
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> http://www.stat.math.ethz.ch/mailman/listinfo/r-help

You are wrong about S-PLUS!  contr.poly only takes as argument the number 
of levels required (and uses pre-computed sets for up to 12 levels).
You get orthogonal polynomials for equally-space values from contr.poly in 
both systems, as documented.

I don't think you are describing contrasts for an ordered factor, but 
orthogonal polynomials in a numeric variable.  The latter are computed by 
the function poly() in both R and S-PLUS.  You could set them up to give a 
contrasts matrix if you want, but not a contrasts function (as that is 
only passed the number of levels, AFAIR).

On Thu, 16 Jan 2003, George W. Gilchrist wrote:
> In S-Plus, I can obtain polynomial contrasts for an ordered factor with
> contr.poly(). The function also exists in R, however is limited to factors
> where the levels are equally spaced. In S-Plus, one can obtain the
contrasts
> for a set of numeric values representing unequally spaced ordered factors.
> Has anyone implemented this in R? I see that the S-Plus function calls
> another function (poly.raw()) that calls a Fortran routine. Thanks for your
> assistance.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595