[R] Dunnett's test
loesljrg at accucom.net
Tue Oct 23 02:17:31 CEST 2001
To: kjetilh at umsanet.edu.bo
Copies to: Kenneth Cabrera <krcabrer at perseus.unalmed.edu.co>,
r-help at stat.math.ethz.ch
Subject: Re: [R] Dunnett's test
From: Peter Dalgaard BSA <p.dalgaard at biostat.ku.dk>
Date sent: 23 Oct 2001 01:03:07 +0200
> kjetil halvorsen <kjetilh at umsanet.edu.bo> writes:
> > > help.search("dunnett")
> > No help files found with alias or title matching `dunnett'
> > so it seems like nobody have implemented this.
> > for how to build it in R, see the " writing R extensions manual", and
> > look in ctest for examples of how to use
> > "htest". For writing dunnett.test, it will be a good idea to return the
> > result as a list with class "htest". Then the printing of the result
> > will be done
> > automatically.
> > Kjetil Halvorsen
> The real obstacle would seem to be the distribution of the many-one
> t-statistic. This doesn't seem to be available from any of the
> standard subroutine sources (apstat and toms). Miller's book cites the
> tables from a 1964 paper by Dunnett - I browsed through it earlier
> today and it seemed that the density can be calculated explicitly, but
> you need numerical integration to get the CDF. Anyone have a handful
> of spare evenings to code this up?
> O__ ---- Peter Dalgaard Blegdamsvej 3
The distribution in question is a certain type of multivariate t.
Dunnett himself provided FORTRAN code for calculating quantiles of
that distribution, in _Applied Statistics_ (1989). [It's Algorithm
AS251.1.]. My copy of the code (downloaded from StatLib, I think)
also has a simple driver/demo program by Dunnett, dated
John R. Gleason
430 Huntington Hall Voice: 315-443-3107
Syracuse, NY 13244-2340 USA FAX: 315-443-4085
PGP public key at keyservers
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch
More information about the R-help