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

RE: Pneumaticity in Triassic pterosaurs

In the early days of numerical taxonomy, the term covered both phenetic and
cladistic-style approaches, although NT has since often been used to refer
mostly to phonetic ones. Phenetic approaches can indeed use continuous data
and, in fact, are most often done with such - specifically very often with
actual measurement data and generating typically Euclidean distance (or
another related metric) matrices which are then subjected to clustering (if
a tree is really wanted - typically the UPGMA clustering algorithm is used
although there are a zillion clustering methods). For binary data there are
fewer options but many of the standard distance/similarity metrics can and
have been used. The simple matching coefficient was a biggie and is
mathematically more pure than others, such as the jaccard or dice, that are
more often used with ecological or biogeographic analyses. 

The distance metric used early with more cladistic related methods was/is
the manhattan or city-block metric. Where Euclidean distance is the direct
distance from A to B, manhattan distance is the distance equivalent of a
taxicab driving in manhattan, with each block change roughly equivalent to a
character change. If there are 3 differences in one variable in four in
another, for example, Euclidean distance will give a distance of 5 and the
manhattan distance will be 3+4=7. There is a nice book for geeks on taxicab
geometry which is fun (like flatland is fun)

It is a great exercise for any evolutionary biologist to experiment with
phenetic and other approaches, as well as clustering, preferably all by
hand. When I've taught classes in data analysis, multivariate analysis, or
ecological ordinations I have had students do this and the good ones leave
really comfortable with the strengths/weaknesses of all methods played with.
I cannot stress strongly enough the importance of NOT doing stuff in a black
box way and, instead, knowing why you take each step.

Probably more than you wanted to know...

Ralph Chapman

-----Original Message-----
From: owner-DINOSAUR@usc.edu [mailto:owner-DINOSAUR@usc.edu] On Behalf Of
Augusto Haro
Sent: Sunday, May 17, 2009 12:03 PM
To: keenir@hotmail.com
Cc: davidpeters@att.net; david.marjanovic@gmx.at; dinosaur@usc.edu
Subject: Re: Pneumaticity in Triassic pterosaurs

As far as I know, you can submit asame matrix with binary characters
to perform either phenetic or phylogenetic algorythms. Phenetic and
phylogenetic analyses are obviously different, based on different
phylosophies and implementing different algorythms, and generally
leading to topologically different trees, but this does not mean that
the same matrices can not be used for both.

I am not aware if in phenetics step-matrix and continuous characters
can be employed, but all these can be made binary, so I guess that
phenetic algorythms can be applied to all data matrices. In case
someone wants to do so, of course, for the utility of phenetics have
been discredited not only in phylogenetics.