Re: Martin 2004 critique (somewhat lengthy)

["Jaime A. Headden" <qilongia@yahoo.com>]

David Marjanovic (david.marjanovic@gmx.at) wrote:

<Parsimony and likelihood methods -- together: cladistics -- don't make any 
sense except in phylogenetics. Neighbor-joining, UPGMA, WPGMA and so on, on the
other hand, are not cladistics -- they are phenetics. It follows that
cladistics is the tool to find a phylogeny.>

  One can apply parsimony without any sort of math, since its an essential
logical argument, and one can apply likelihood and its many maths, without ever
trodding on the field of phylogeny recovery. That they are used to analyze data
to produce phylogeny is only a logical output of their existence.

  Some sources of data on the subject include -- but are not limited to -- the
following:

  http://www.tk421.net/essays/simple.html

  Likelihood: "Likelihood is the hypothetical probability that an event that
has already occurred would yield a specific outcome. The concept differs from
that of a probability in that a probability refers to the occurrence of future
events, while a likelihood refers to past events with known outcomes." From
Mathworld: http://mathworld.wolfram.com/Likelihood.html

  http://mathworld.wolfram.com/MaximumLikelihood.html

  Phylogeny doesn't really appear in any of these. The prime example of
parsimony, in fact, is Occam's Razor, in which the principle has been applied
throughout the sciences, and indeed by any scientific-minded fellow to his
observations, as a means of testing hypotheses. Thus it has little core
application, even with the math of likelihood involved, toward phylogeny.

  As I also stated before, it's possible but often ignored to use parsimony or
likelihood methods to test datasets that are designed to produce a phylogenetic
hypothesis, but applied to non-phylogenetic criteria. Examples include studying
gene recombination, gene family origins (which often require NO phylogeny in
testing bacterial lateral transfer to study how recombination events operate in
steps to produce new genes without any phylogenetic "splitting" occuring), even
languages, which operate much the same way above. They can also be used to test
absolute phenetic similarity, without any phylogenetic inferrence involved, but
which may then be applied phylogenetically. This is part of what I am working
on. One can make reference from this work that there is a phylogenetic
relationship observable in the data, but this is a separate process from the
studies performed themselves, and indeed the math and machines used do not care
what data is applied as, they just crunch the numbers.

  As I stated in my first reply, was that cladistics is the study of how parts
of data sets are compared to other parts of the same data sets, and that is all
these programs are capable of doing. Any phylogenetic extension of this data is
secondary and external to the programming.

  Further in his reply, David seems to have misconstrued an innate similarity
between phylogeny and phenetics. While they seem related, it should be noted
that shared descent may utilize an observable phenetic paradigm, but the
phenetics is separate from the association of commonj descent. 

  Some other thoughts here:

  http://www.icp.ucl.ac.be/~opperd/private/phenetics.html

Quoting from above:
  "The maximum parsimony method is a typical representative of the cladistic
   approach, whereas the UPGMA method is a typical phenetic method. The other
   methods, however, cannot be classified easily according to the above
   criteria. For example, the transformed distance method and the neighbors
   relation method have often been said to be phenetic methods, but this is not
   an accurate description. Although these methods use similarity (or
   dissimilarity, i.e., distance) measures, they do not assume a direct
   connection between similarity and evolutionary relationship, nor are they
   intended to infer phenetic relationships."

  Indeed, one can observe the similarities among cars and arrive at a cladogram
of phenetic similarity among all car makes ever made and aquire a "tree" that
has nothing at all to do with phylogeny (cars don't breed) which may even group
most cars by their makes or even years (the Nissan Z series, perhaps, or Ford's
early models). What this output is not, however, is an hypothesis of common
derivation, as this theory is applied after or before the data is ran, and thus
the data itself cares little for the subjective reasoning in the observer. This
messes with the test, but all it is is a phenetic diagram. Inferring time into
it makes it very, very different. I can, for example, create a phenetic diagram
by grouping colors by their frequencies, and arrive at a (theoretically)
sequentially nested phyllogram, or a "bush", which relates only to closeness of
the numbers to one another. Such is the same with gene trees.

  Hopefully what I have written may help illustrate the reducible aspects of
phylogenetic reconstruction and the ability to apply several tools in this
science for the sake of arriving at various different studies, including whole
phenetic similarity versus a recovered phylogeny. And I hope that the actual
difference between parts of the whole and the whole from its parts helps create
a healthy distrust of results via the process the data goes through, as one
learns to differentiate the parts as useful tools on their own. One may note I
still use these methods as a whole, as in PAUP* to test phylogenies, so there's
hope for me yet!

  Cheers,

Jaime A. Headden

  Little steps are often the hardest to take.  We are too used to making leaps 
in the face of adversity, that a simple skip is so hard to do.  We should all 
learn to walk soft, walk small, see the world around us rather than zoom by it.

"Innocent, unbiased observation is a myth." --- P.B. Medawar (1969)


                
__________________________________ 
Discover Yahoo! 
Use Yahoo! to plan a weekend, have fun online and more. Check it out! 
http://discover.yahoo.com/