Striking a balance, weighting and Cladistics

Thomas DiBenedetto TDibenedetto at DCCMC.ORG
Wed Feb 21 09:40:07 CST 2001


-----Original Message-----
From: Richard Zander [mailto:rzander at SCIENCEBUFF.ORG]
Tom:
You cling tenaciously to the idea that a single tree or a bunch of equally
parsimonious trees must be a "result" or a "preference." I suggest that only
when the output of a computer program has demonstrably high support is an
admissible as a "result" or a "preference." It is otherwise nothing.

Richard,
        When I use the terms "result" or "preference", I am simply referring
to the output of the analysis, without any implication that the result is
necessarily meaningful. I recognize that the terms can be qualified by
saying (e.g.) "meaningless result" or "trivial result" or "weakly supported
preference". As I stated previously, standard cladistic practices include
some metric of "strength of support" along with the "result" - the set of
equally most-parsimonious trees. If nodes exhibit a high decay index, or
high bootstrap or jackknife numbers, and the analysis has included all
available evidence relevant to the taxa, then one can have confidence that
the result represents our best reconstruction of the phylogeny, and we can
proceed to use that result in further evolutionary studies, and to make
changes in classification, if appropriate. But if the strength numbers are
low, or the data is generated from only a subset of the available evidence,
then such further steps are not warrented.
In brief, I agree completely with your underlying concerns, but I disagree
with your assumption that these concerns are not addressed by standard
practices.
-tom




More information about the Taxacom mailing list