More taxo-computational distractions

[Richard Zander]

homas Schlemmermeyer wrote:
>
> Dear all,
>
> my study of phylogenetics is now already advancing. But, before jumping from
> taxa to com, I would still like to continue pondering about the basic
> underlying philosophical principles.
>
> So, I started reading "Reconstructing the past - parsimony, evolution
> and inference" by Elliott Sober
> (1988 Massachusetts Institute of technology).
> It is all about the question wether parsimony assumptions make sense at all.
> SO, here is my question: Do they make sense?


Yes. They make sense in that maximum synapomorphy parsimoniously
eliminates grossly unreasonable trees, and maximum likelihood eliminates
grossly improbable gene trees. With both, however, there remains a pool
of Darwinianly reasonable trees.

No. They don't make sense in that identifying the tree of maximum
synapomorphy as the best tree ignores the pool of tens or hundreds of
trees that are slightly more falsifiable. And maximum likelihoods that
do not involve posterior probabilities greater than .5 do not present
probabilistic hypotheses; there are usually many other trees whose sum
of posterior probabilities is higher than that of the tree of maximum
likelihood.

This all revolves around misapplication of the theory "statistical
relevance" in phylogenetics. In medicine, a person who has taken a test
that increases the chance of a disease to, say, 1 in 20, when the
population has only a 1 in 10000 chance, is of course counseled to take
further tests. The increase matters. The same is true in all cases where
a single event may have a chance lower than .5 but an increased
probability, AND the risk is great. This is not appropriate in
phylogenetic studies where the chance of a hypothetical tree being the
true tree is greatly increased but still less than .5. This is because
there remains a pool of trees deserving the same attention as possible
"best" hypotheses.

I have a paper on this in press. I'll put you on my mailing list.

R.
--

*******************************************************
Richard H. Zander, Buffalo Museum of Science
1020 Humboldt Pkwy, Buffalo, NY 14211 USA bryo at commtech.net
*******************************************************