[Fwd: Re: Probabilities on Phylogenetic Trees]

James Francis Lyons-Weiler weiler at ERS.UNR.EDU
Fri Sep 19 07:38:54 CDT 1997

Look, on my criticisms of parsimony, Tom, you hearken constantly back to
questions about whether ANY method will be able to extract the signal that
is, in fact, present, and then you attempt to place me in the category of
people who think they can accurately model evolutionary processes instead.
My criticisms stand independent of whether an alternative methods for
stydying hypotheses of homology and getting trees exists...

Parsimony can get a leg up in the long run if such problematic data sets
are avoided (a la signal testing), but then so can other methods.  The
question of whether the processes of evolution have, for matrix A, left
behind patterns that we expect to lead to decent evolutionary trees is to
some degree independent of which method of tree estimation one adopts.

On bushes = longest tree; that's an interesting twist, in part because I
was under the illusion we were talking about resolved hypotheses.  I
consider the bush topology in every case to represent something of a null
model, but there are other null hypotheses that can be adopted, and other
response criteria than tree length that can be tested to ask questions
about the amount of different types of structure present in a matrix.

For all the volleys, and semantic difficulties, and differential focus on
trees vs. hypotheses of homology, there are remarkably few differences
among people (minus the pattern cladists, which you obviously do not
identify with) with respect to purpose and interest.  Few people, in my
opinion, take their favorite method of inference to task.  I see value in
stating the pitfalls, and being up front about limitations of methods as
they are currently understood, and in trying to find where the limits of
methods in fact exist.  The fact that parsimony may be just as bad as any
other method under some conditions is no relief.  The fact that the
accuracy of trees can be improved by (1) avoiding noisy data without
modelling evolution, (2) detecting and mitigating the influences of long
branches when they exist, and (3) finding out if adding more data will
make things better or worse (these are parts of the horse and pony show
for rasa) will, in my best of all possible worlds, make the data speak
more clearly when it is queried with the parsimony algorithm.

All of this got started with Richard Zander asking the question, "what
effect do (1) the data, (2) the algorithm, (3) the criterion, and (4) the
researcher have on the P(t = T)?  I'd like to suggest that we pull on that
thread again for a while without necessarily going back to our respective
corners and waiting for the next bell.  I find that question new, fresh,
relevant and worthy of some hard thought.

(I'm not bored... just busy.)



  \  /   /    \  /           JAMES LYONS-WEILER           ______________
   \/   /      \/                                        |..............|
    \  /       /                                         |..............|
     \/       /              DOCTORAL PROGRAM IN         |..............|
      \      /               ECOLOGY, EVOLUTION, AND     |...***........|
       \    /                CONSERVATION BIOLOGY        |..*****.......|
        \  /                                             |.******.......|
         \/                  1000 VALLEY ROAD/186        |********......|
    ______________           THE UNIVERSITY OF            --------------
   | will perform |          NEVADA, RENO
   |  statistical |          RENO, NEVADA 89512-0013
   | phylogenetic |
   | analyses for |         "(Biology) is not religion; if it were, we'd
   |    food      |          have a much easier time raising money."
    --------------                       -Leon Lederman

                RASA 2.1 SOFTWARE FOR THE MAC
                Download @

More information about the Taxacom mailing list