Hi Karin.
GLS with x as a factor is a generalized ANOVA which assumes [in the case
of gls(...,correlation=corBrownian)] that the residual error in the
ANOVA model has evolved by Brownian evolution. If you read your data
into data frame Z with row names as species names, for instance:
Z<-read.table("filename",header=T,row.names=1)
tree<-read.tree("treefile")
and your column name for the factor is x & the column name for the
continuous response variable is y, then you should just be able to do:
fit<-gls(y~x,data=Z,correlation=corBrownian(1,tree))
You can then perform various posthoc analyses from the gls object that
is produced. For instance
summary(fit)
anova(fit)
residuals(fit)
As pointed out by Alejandro, you should check for normality of the
residuals in residuals(fit) - not the normality of y before analysis.
summary(fit) will also give you parameter estimated (fitted means for
each factor) and standard errors. These can be used to conduct posthoc
comparison of means using t-tests in the standard way.
I hope this helps.
All the best, Liam
--
Liam J. Revell
University of Massachusetts Boston
web: http://faculty.umb.edu/liam.revell/
email: liam.rev...@umb.edu
blog: http://phytools.blogspot.com
On 5/30/2012 10:46 AM, Karin Schneeberger wrote:
Hi Alejandro
Thank you for the very quick answer. I tried PGLS before, but then was told
that GLS is not suitable for multistate categorical variables that can not be
ranked (otherwise I would treat them as continuous). Also, with GLS it's as far
as I understood not possible to state statistically whether certain groups are
greater than others. But I am new into this kind of analysis and am very happy
for any help and explanation, as I might be totally wrong.
Cheers,
Karin
________________________________
Von: Alejandro Gonzalez<alejandro.gonza...@ebd.csic.es>
CC: "r-sig-phylo@r-project.org"<r-sig-phylo@r-project.org>
Gesendet: 16:26 Mittwoch, 30.Mai 2012
Betreff: Re: [R-sig-phylo] Non-parametric alternative to phylogenetic ANOVA?
Hi Karin,
You could use a gls method and look at the distribution of your residuals. It
is the residuals which must be normally distributed, which can be checked using
diagnostic plots such as a histogram or qq-plot of the residuals of your model.
Cheers
Alejandro
On 30, May 2012, at 4:12 PM, Karin Schneeberger wrote:
Dear all
I'm trying to compare one trait across three (unordered categorical) groups
including 25 species (let's say for example basal metabolic rate of aquatic,
terrestrial and aerial mammals).
If the data would be normally distributed, I would simply use a phylogenetic ANOVA
including a post-hoc test on means accounting for the phylogeny, as provided by the
package "phytools". However, my tip-data are far away from being normally
distributed, and transformations only lead to unsatisfying improvements (e.g Shapiro-Wilk
test returns a p-Value of 0.08 instead of 0.03 as before transformation). So I am not
convinced that a phylogenetic ANOVA is the right approach for dealing with these sort of
data.
Is there any non-parametric approach to compare groups across phylogeny that
also returns which groups differ from each other?
Your sincerely,
Karin
[[alternative HTML version deleted]]
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
__________________________________
Alejandro Gonzalez Voyer
Post-doc
Estación Biológica de Doñana
Consejo Superior de Investigaciones Científicas (CSIC)
Av Américo Vespucio s/n
41092 Sevilla
Spain
Tel: + 34 - 954 466700, ext 1749
E-mail: alejandro.gonza...@ebd.csic.es
Web site (Under construction):
Personal page: http://consevol.org/members/alejandro.html
Group page: http://consevol.org/index.html
For PDF copies of papers see:
http://csic.academia.edu/AlejandroGonzalezVoyer
[[alternative HTML version deleted]]
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo