[R] R vs SPSS contrasts

[Gabor Grothendieck]

The formula should be (diag(n) - 1/n)[, -n]

On Sun, Oct 12, 2008 at 1:36 PM, Gabor Grothendieck
<ggrothendieck at gmail.com> wrote:
> Looks like the contrast matrix for indicator is contr.SAS(n),
> for deviation is contr.sum(n) and for simple is:
>
> (diag(n) - 1/n)[, -1]
>
> That works at least for the n = 3 example in the link.
> Perhaps the others could be checked against SPSS
> for a variety of values of n to be sure.
>
> On Sun, Oct 12, 2008 at 12:32 PM, Chuck Cleland <ccleland at optonline.net> wrote:
>> On 10/11/2008 3:31 PM, Ted Harding wrote:
>>> Hi Folks,
>>>
>>> I'm comparing some output from R with output from SPSS.
>>> The coefficients of the independent variables (which are
>>> all factors, each at 2 levels) are identical.
>>>
>>> However, R's Intercept (using default contr.treatment)
>>> differs from SPSS's 'constant'. It seems that the contrasts
>>> were set in SPSS using
>>>
>>>   /CONTRAST (varname)=Simple(1)
>>>
>>> I can get R's Intercept to match SPSS's 'constant' if I use
>>> contr.sum in R.
>>>
>>> Can someone please confirm that that is a correct match for
>>> the SPSS "Simple(1)", with identical effect?
>>>
>>> And is there a convenient on-line reference where I can look
>>> up what SPSS's "/CONTRAST" statements exactly mean?
>>> I've done a lot of googling, withbout coming up with anything
>>> satisfactory.
>>>
>>> With thanks,
>>> Ted.
>>
>> Hi Ted:
>>  Here are two links with the same content giving a brief description of
>> SPSS simple contrasts:
>>
>> http://www.ats.ucla.edu/stat/spss/library/contrast.htm
>> http://support.spss.com/productsext/spss/documentation/statistics/articles/contrast.htm
>>
>>  These pages explain how simple contrasts differ from indicator
>> (contr.treatment) and deviation (contr.sum) contrasts.  For a factor
>> with 3 levels, I believe you can reproduce SPSS simple contrasts (with
>> the first category as reference) like this:
>>
>>> C(warpbreaks$tension, contr=matrix(c(-1/3,2/3,-1/3,-1/3,-1/3,2/3),
>> ncol=2))
>> ...
>> attr(,"contrasts")
>>        [,1]       [,2]
>> L -0.3333333 -0.3333333
>> M  0.6666667 -0.3333333
>> H -0.3333333  0.6666667
>> Levels: L M H
>>
>>  For a factor with 2 levels, like this:
>>
>>> C(warpbreaks$wool, contr=matrix(c(-1/2,1/2), ncol=1))
>> ...
>> attr(,"contrasts")
>>  [,1]
>> A -0.5
>> B  0.5
>> Levels: A B
>>
>>  Your description of the effect of SPSS simple contrasts - intercept
>> coefficient of contr.sum and non-intercept coefficients of
>> contr.treatment - sounds accurate to me.
>>
>> hope this helps,
>>
>> Chuck
>>
>>> --------------------------------------------------------------------
>>> E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk>
>>> Fax-to-email: +44 (0)870 094 0861
>>> Date: 11-Oct-08                                       Time: 20:31:53
>>> ------------------------------ XFMail ------------------------------
>>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list
>>> 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.
>>
>> --
>> Chuck Cleland, Ph.D.
>> NDRI, Inc. (www.ndri.org)
>> 71 West 23rd Street, 8th floor
>> New York, NY 10010
>> tel: (212) 845-4495 (Tu, Th)
>> tel: (732) 512-0171 (M, W, F)
>> fax: (917) 438-0894
>>
>> ______________________________________________
>> R-help at r-project.org mailing list
>> 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.
>>
>