[R] What are Type II or III contrast? (contrast() in contrast package)

[Emmanuel Charpentier]

Le mercredi 03 février 2010 à 09:23 -0600, Peng Yu a écrit :
> On Wed, Feb 3, 2010 at 2:12 AM, Emmanuel Charpentier
> <charpent at bacbuc.dyndns.org> wrote:
> > Le mercredi 03 février 2010 à 00:01 -0500, David Winsemius a écrit :
> >> On Feb 2, 2010, at 11:38 PM, Peng Yu wrote:
> >>
> >> > ?contrast in the contrast package gives me the following description.
> >> > However, I have no idea what Type II and III contrasts are. Could
> >> > somebody explain it to me? And what does 'type' mean here?
> >> >
> >> >    *‘type’*: set ‘type="average"’ to average the individual contrasts
> >> >    (e.g., to obtain a Type II or III contrast)
> >>
> >> In no particular order:
> >> http://courses.washington.edu/b570/handouts/type3ss.pdf
> >> http://core.ecu.edu/psyc/wuenschk/SAS/SS1234.doc
> >> http://n4.nabble.com/a-kinder-view-of-Type-III-SS-td847282.html
> >>
> >> Don't expect any follow-up questions to be answered or further
> >> citations offered. This is really more in the realm of statistics
> >> education than an R specific question.
> >
> > Nonwhistanding David Winsemius' closing remark, I'd like to add
> > something that should be requested reading (and maybe hinted at in
> > lm()'s help page) :
> >
> > http://www.stats.ox.ac.uk/pub/MASS3/Exegeses.pdf
> >
> > (BTW, despite is age, MASS *is* requested reading, and Bill Venables'
> > exegeses should be part of it).
> 
> Do you by any means indicate that MASS describes Type I, II, III, IV
> contrast? Although MASS describes contrasts, but I don't think it
> describes Type I, II, III, IV contrast.

it does not. But it explains (tersely) *WHY* one has *serious* logical
difficulties when tryng to interpret contrasts not abiding to
marginality principle.

HTH,

					Emmanuel Charpentier