[R] multiple comparisons/tukey kramer

Tyler Smith tyler.smith at mail.mcgill.ca
Fri Nov 23 22:49:24 CET 2007


Thank you for your response. I think you have misunderstood what I'm
asking, though.

On 2007-11-23, Emmanuel Charpentier <charpent at bacbuc.dyndns.org> wrote:
>
> - Tukey HSD will enable you to test the p(p-1)/2 pair differences one
> can create with p groups ;
> - Dunnett's procedure is made to compare (p-1) "treatments" to a common
> control ;
> - Scheffé's procedure is applicable to *any* ("reasonable") set of
> contrasts you can form ;
> - Newman-Keuls : aims to create separate subset of groups (but has
> serious conceptual and technical flaws ! Don't do that nunless you know
> what you're doing...).
> - etc ...
>
> You'll have to refer to the subject matter to make a choice.

Of course. I also have to know which function in R corresponds to which
test, which is my main question.

> Google ("multiple comparisons") will offer you some dubious and quite a
> few good references...

I have indeed found many dubious and a few good references to multiple
comparisons, both from google and r-site-search. Many posts in the
archive, including one made today in response to another question of
mine, point to glht as the appropriate function to use in R.

However, I don't know what exactly glht does, and the help file is
extremely terse. It offers the following options (in contrMat()):

     contrMat(n, type=c("Dunnett", "Tukey", "Sequen", "AVE", 
                        "Changepoint", "Williams", "Marcus", 
                        "McDermott"), base = 1)

The only reference to the source of these tests is:

     Frank Bretz, Alan Genz and Ludwig A. Hothorn (2001), On the
     numerical  availability of multiple comparison procedures.
     _Biometrical Journal_, *43*(5), 645-656.

This is a very technical paper, which as far as I can follow, is
primarily a discussion of the numerical methods involved in
calculating these contrasts, rather than the contrasts themselves. I
can't decide which one is appropriate without knowing what the
differences are. Dunnett seems pretty straightforward. Tukey, I think,
may refer to what is referred to as the Tukey-Kramer test in other
sources? Are any of them related to Scheffe? I have no idea. None of
them are related to Newman-Keuls, as several archive messages make
very clear that this is not a valid comparison to use, so R doesn't
implement it.

What I need is a reference to the tests implemented in glht, so I can
decide which one is appropriate for my data. Sequen, Changepoint et
al. may be common terms in some fields, but not in the references I'm
working from.

Thanks,

Tyler



More information about the R-help mailing list