[R] adjusted p-values with TukeyHSD?

Liaw, Andy andy_liaw at merck.com
Tue May 17 15:19:48 CEST 2005


> From: Sander Oom
> 
> Hi Chris and Chris,
> 
> I was keeping my eye on this thread as I have also been discovering 
> multiple comparisons recently. Your instructions are very 
> clear! Thanks.

One thing to note, though:  Multcomp does not do Dunnett's or 
Tukey's multiple comparisons per se.  Those names in multcomp 
refer to the contrasts being used (comparison to a control for 
Dunnett and all pairwise comparison for Tukey).  The actual 
methods used are as described in the references of the help
pages.

 
> Now I would love to see an R boffin write a nifty function to 
> produce a 
> graphical representation of the multiple comparison, like this one:
> 
> http://www.theses.ulaval.ca/2003/21026/21026024.jpg
> 
> Should not be too difficult.....[any one up for the challenge?]

I beg to differ:  That's probably as bad a way as one can use to 
graphically show multiple comparison.  The shaded bars serve no 
purpose.

Two alternatives that I'm aware of are 

- Multiple comparison circles, due to John Sall, and not 
  surprisingly, implemented in JMP and SAS/Insight.  See:
 
http://support.sas.com/documentation/onlinedoc/v7/whatsnew/insight/sect4.htm


- The mean-mean display proposed by Hsu and Peruggia:
  Hsu, J. C. and M. Peruggia (1994). 
  Graphical representations of Tukey's multiple comparison method.
  Journal of Computational and Graphical Statistics 3, 143{161

Andy
 
> I came across more multiple comparison info here;
> 
> http://www.agr.kuleuven.ac.be/vakken/statisticsbyR/ANOVAbyRr/m
> ultiplecomp.htm
> 
> Cheers,
> 
> Sander.
> 
> Christoph Buser wrote:
> > Dear Christoph
> > 
> > You can use the multcomp package. Please have a look at the
> > following example:
> > 
> > library(multcomp)
> > 
> > The first two lines were already proposed by Erin Hodgess:
> > 
> > summary(fm1 <- aov(breaks ~ wool + tension, data = warpbreaks))
> > TukeyHSD(fm1, "tension", ordered = TRUE)
> > 
> >     Tukey multiple comparisons of means
> >     95% family-wise confidence level
> >     factor levels have been ordered
> >  
> > Fit: aov(formula = breaks ~ wool + tension, data = warpbreaks)
> > 
> > $tension
> >          diff        lwr      upr
> > M-H  4.722222 -4.6311985 14.07564
> > L-H 14.722222  5.3688015 24.07564
> > L-M 10.000000  0.6465793 19.35342
> >  
> > 
> > By using the functions simtest or simint you can get the
> > p-values, too:
> > 
> > summary(simtest(breaks ~ wool + tension, data = warpbreaks, 
> whichf="tension",
> >         type = "Tukey"))
> > 
> > 	 Simultaneous tests: Tukey contrasts 
> > 
> > Call: 
> > simtest.formula(formula = breaks ~ wool + tension, data = 
> warpbreaks, 
> >     whichf = "tension", type = "Tukey")
> > 
> > 	 Tukey contrasts for factor tension, covariable:  wool 
> > 
> > Contrast matrix:
> >                       tensionL tensionM tensionH
> > tensionM-tensionL 0 0       -1        1        0
> > tensionH-tensionL 0 0       -1        0        1
> > tensionH-tensionM 0 0        0       -1        1
> > 
> > 
> > Absolute Error Tolerance:  0.001 
> > 
> > Coefficients:
> >                   Estimate t value Std.Err. p raw p Bonf p adj
> > tensionH-tensionL  -14.722  -3.802    3.872 0.000  0.001 0.001
> > tensionM-tensionL  -10.000  -2.582    3.872 0.013  0.026 0.024
> > tensionH-tensionM   -4.722  -1.219    3.872 0.228  0.228 0.228
> > 
> > 
> > 
> > or if you prefer to get the confidence intervals, too, you can
> > use:
> > 
> > summary(simint(breaks ~ wool + tension, data = warpbreaks, 
> whichf="tension",
> >         type = "Tukey"))
> > 
> > 	Simultaneous 95% confidence intervals: Tukey contrasts
> > 
> > Call: 
> > simint.formula(formula = breaks ~ wool + tension, data = 
> warpbreaks, 
> >     whichf = "tension", type = "Tukey")
> > 
> > 	 Tukey contrasts for factor tension, covariable:  wool 
> > 
> > Contrast matrix:
> >                       tensionL tensionM tensionH
> > tensionM-tensionL 0 0       -1        1        0
> > tensionH-tensionL 0 0       -1        0        1
> > tensionH-tensionM 0 0        0       -1        1
> > 
> > Absolute Error Tolerance:  0.001 
> > 
> >  95 % quantile:  2.415 
> > 
> > Coefficients:
> >                   Estimate   2.5 % 97.5 % t value Std.Err. 
> p raw p Bonf p adj
> > tensionM-tensionL  -10.000 -19.352 -0.648  -2.582    3.872 
> 0.013  0.038 0.034
> > tensionH-tensionL  -14.722 -24.074 -5.370  -3.802    3.872 
> 0.000  0.001 0.001
> > tensionH-tensionM   -4.722 -14.074  4.630  -1.219    3.872 
> 0.228  0.685 0.447
> > 
> > -----------------------------------------------------------------
> > Please be careful: The resulting confidence intervals in
> > simint are not associated with the p-values from 'simtest' as it
> > is described in the help page of the two functions.
> > -----------------------------------------------------------------
> > 
> > I had not the time to check the differences in the function or
> > read the references given on the help page.
> > If you are interested in the function you can check those to
> > find out which one you prefer.
> > 
> > Best regards,
> > 
> > Christoph Buser
> > 
> > --------------------------------------------------------------
> > Christoph Buser <buser at stat.math.ethz.ch>
> > Seminar fuer Statistik, LEO C13
> > ETH (Federal Inst. Technology)	8092 Zurich	 SWITZERLAND
> > phone: x-41-44-632-4673		fax: 632-1228
> > http://stat.ethz.ch/~buser/
> > --------------------------------------------------------------
> > 
> > 
> > Christoph Strehblow writes:
> >  > hi list,
> >  > 
> >  > i have to ask you again, having tried and searched for 
> several days...
> >  > 
> >  > i want to do a TukeyHSD after an Anova, and want to get 
> the adjusted  
> >  > p-values after the Tukey Correction.
> >  > i found the p.adjust function, but it can only correct 
> for "holm",  
> >  > "hochberg", bonferroni", but not "Tukey".
> >  > 
> >  > Is it not possbile to get adjusted p-values after 
> Tukey-correction?
> >  > 
> >  > sorry, if this is an often-answered-question, but i 
> didn´t find it on  
> >  > the list archive...
> >  > 
> >  > thx a lot, list, Chris
> >  > 
> >  > 
> >  > Christoph Strehblow, MD
> >  > Department of Rheumatology, Diabetes and Endocrinology
> >  > Wilhelminenspital, Vienna, Austria
> >  > chrisxe at gmx.at
> >  > 
> >  > ______________________________________________
> >  > R-help at stat.math.ethz.ch mailing list
> >  > https://stat.ethz.ch/mailman/listinfo/r-help
> >  > PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html
> > 
> > 
> ______________________________________________
> > R-help at stat.math.ethz.ch mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html
> > 
> 
> 
> -- 
> 
> --------------------------------------------
> Dr Sander P. Oom
> Animal, Plant and Environmental Sciences,
> University of the Witwatersrand
> Private Bag 3, Wits 2050, South Africa
> Tel (work)      +27 (0)11 717 64 04
> Tel (home)      +27 (0)18 297 44 51
> Fax             +27 (0)18 299 24 64
> Email   sander at oomvanlieshout.net
> Web     www.oomvanlieshout.net/sander
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html
> 
> 
>




More information about the R-help mailing list