Joe Felsenstein wrote on 26/09/2009 21:48:
Elsa et Stéphane Bouee wrote:
I am currently doing a classification analysis on morphologic data using
modern methods of morphometry (procruste analysis).
The variables obtained are quantitative and I would like to use them in a
phylogenetic approach.
For this purpose I used 2 methods:
1) maximum likelihood with the software phylips and its contml
add on;
As the author of Contml, which is a program, not an "add-on" (whatever that
means) I can perhaps comment.
2) cladistic analysis on quantitative variables, with the
software TNT
developed by Goloboff, Farris and Nixon.
I am wondering if the max likelihood is really performing a phylogenetic
classification or if it is rather a kind of phenetic method. Maybe the
answer may vary according to whom I ask the question …?
I have submitted an article presenting the max likelihood as a
phylogenetic
classification but the reviewers are challenging this assertion.
Good luck with the reviewers -- I have never been able to influence
people who had strong views about parsimony being better than everything
else.
There's nothing wrong with the Procrustes superposition. But the
assumptions
of Contml are that each coordinate is independently changing according to
a Brownian motion process, and all at the same rate.
This is a dubious assumption at best. I have written about this and warned
about this for years (Amer J Human Genetics 1973; Evolution 1981; Annual
Review of Ecology and Systematics 1988; article in volume on Morphology,
Shape,
and Phylogenetics 2002). I have an extensive discussion of the issue in my
2004 book, chapter 24. In PHYLIP the documentation file contchars.html
discusses it too. Ideally it would be best to have some estimate of
the covariation between the characters, and use that to transform your
coordinates
to independence. I don't know an easy way to do that unless you have a
known phylogeny and can estimate the covariances along it using comparative
methods software.
So the referees are right to be skeptical, but they should be equally
skeptical of
discrete characters parsimony approaches which make quite similar
assumptions. The hardest-core cladistics people do not admit this.
As what classifications should be, or whether methods should be
considered as making phylogenetic or phenetic classifications, I have my
own position, that no one else seems to back (in public, anyway). I
think that we
should not think of these trees as classifications, and not call them
phylogenetic
classifications or phenetic classifications, but consider them as
estimates of the
phylogeny. The issue of how to classify is less important anyway.
I have the strong feeling that most users of R and its [phylo]genetics
packages are interested in the study of evolutionary processes, not in
classification (I rarely see questions about classification or
systematics here). So maybe most of us silently back Joe's position.
About the issue of how to classify, I think it is very important. The
point here is, in my view, that the confusion between classification and
evolution greatly hampered the progress of evolutionary biology, but the
situation has improved in recent years.
Emmanuel
J.F.
----
Joe Felsenstein, j...@gs.washington.edu
Dept. of Genome Sciences, Univ. of Washington
Box 355065, Seattle, WA 98195-5065 USA
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
--
Emmanuel Paradis
IRD, Montpellier, France
ph: +33 (0)4 67 16 64 47
fax: +33 (0)4 67 16 64 40
http://ape.mpl.ird.fr/
_______________________________________________
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
