[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Benton et al.'s Supertree (long)



> No, the use of multiple trees with the same data implies that 
> historically earlier discovered characters will be overweighted in 
> the analysis.  This is not a defendable reason for weighting 
> characters.  Any "consensus" seen due
> to this is caused by a lack of new data or ideas until reexamination
> or new specimens indicated an alternate phylogeny.

The "weighting" I mentioned is not the weighting of characters, it is
the weighting of sources.  For example, it is common practice to take
all of the related trees from one paper, combine them into a small
tree, and then put them in the supertree together, thus giving the
group a single unit instead of overweighting that study.

> How is valuing someone's undefended opinion of a relationship
equally to
> someone else's result of a 300+ character phylogenetic analysis  
> useful?  ... An earlier tree with less
> characters and taxa available can only be "bad" compared to a modern
> analysis.

Yes, the equal weighting has its weaknesses.  However, in most cases,
it is not clear whether one tree is better than another.  A more
common situation would be a 70 character tree versus a 65 character
tree.  Yes, one has more characters than the other, but are you really
going to say the 70 character has to be better because it has 5 more
traits in the matrix?  Things also get messy when you try to work out
what a "good" trait is.  These types of decisions and debates
certainly are important in phylogenetics, and equal weighting is one
way to approach it.

That being said, I would agree that repetition of data from the SAME
AUTHOR(s) is troubling.  It is one thing to use a tree from a paper
that is partially adopted from another, since this shows that the
later paper agrees with the earlier analysis (othewise they wouldn't
use it), and this is useful information.  However, people usually
agree with themselves (there are notable exceptions), and thus the
same data from the same person is probably dangerous repetition.  I
agree with you that Benton should have come up with an alternative in
these cases.

The easiest way to determine how much this affects the tree would be
to redo the analysis without this repetition, and see whether the
topology changes significantly.  If not, then it is a moot point (at
least for that particular tree).
  
> > Finally, supertrees give the result of using
> > several types of characters/phylogenetic information.
> 
> They CAN, but Benton et al.'s doesn't.  It's all morphological.  As
long as
> the characters in each included analysis are separate, I have no
problem
> with supertrees.  If you made a supertree using separate 12S, 16S,
> behavioral and morphological data sets, it would be fine.  Benton et
al.'s
> however, does not have this advantage.

Ah, but different morphological characters are used by different
analyses.  I agree that Benton's tree is weaker than others to some
extent in that it lacks molecular and behavioral data (nothing he can
do about that, of course).  However, there are still different types
of analysis going into the tree in the sense that different aspects of
morphology are used in different studies.

One major weakness of the current supertree from Benton et al is the
lack of meaningful branch lengths.  This means that no analysis
requiring rates can be done, and this kills a fair bit of the point
(in my opinion).  I suppose they may try to do this in the future, as
another publication, but I would tend to think it would be difficult
for this tree.

--Mike Habib