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

Re: Learning cladistics (was Re: Dinosaur Web Pages' Re-Opening)

At 08:48 PM 9/5/97 -0500, Jonathan R. Wagner wrote:
>... Given that, the explanation which requires the least numer of
>unsubstantiated assumptions, and the fewest untestable hypotheses,
>is most likely to be the correct one.

True, but the question is, are character transitions the only hypotheses
implied by a cladistic analysis?  I think not, it also implies things like
ghost lineages, dichotomies, and maybe other things as well.  True
parsimony would require that these also be included in the parsimony
analysis.  (For instance, *real* evolution almost certainly is NOT
restricted to just dichotomous branching - species tend to bud, leading to
polychotomies; and the parent species may easily be polymorphous for traits
that become fixed in descendents).

However my real beef is with the lack of statistical testing.  When large
numbers of characters and/or transitions are involved a difference of two
or three steps simply is *not* going to be statistically significant, IMHO.
 That is the several "nearly most parsimonious" trees may be effectively as
likely as the most parsimonious one.  Statistically, one should only treat
a tree as less likely that the most parsimonious one if the difference in
number of steps is statistically significant, at least at the 95%
confidence level.

There are apparently statistical tests available for DNA based cladograms,
but not yet for character based ones.  This is a sore need.  One of the
recently described DNA statistical tests might be adaptable to use with
character data, but applying it would require massive amounts of computation.

> And given that assumption, the application of
>the principle of parsimony should only fail to produce the best 
>hypothesis in the absense of some data.

Missing species in the analysis.

Missing characters in known species

Lack of data on polymorphisms.

Hmm, there generally seems to be lots of missing data.

> However, you cannot simply say that, because of the lack
>of data, all cladistics is bunk, or we should disregard cladistic 
>analyses, even when backed up by non-cladistic evidence, ...

I generally do accept cladistic analyses as a generally adequate model.  (I
am not quite as iconoclastic as George Olshevsky in this regard).

My main diffidence is in accepting weakly supported clades as real.  I tend
to treat them as an over-resolved polychotomy.  (Over-resolved because
cladistic analysis as now performed pretty much forces dichotomies, even
where multiple budding from a polymorphous species is actually more likely
- that alternative is not even considered, so its relative parsimony is not
easily obtainable).

Hmm, this suggests comparing the number of steps implied by a polychotomy,
where any character change that appears at the base of more than one branch
is treated as a *single* character change.

May the peace of God be with you.         sarima@ix.netcom.com