parsimony/biology

Zden�k Sk�la skala at INCOMA.CZ
Tue Feb 27 11:39:11 CST 2001


>>...which source of information you use to distinguish homology from
>>non-homology?
>Your knowledge of biology provides the initital tests. Congruence, an
>expectation of theory, is another test.
Well, seems OK to me - my only point was that the "character-matrix
homology" (before test) and "cladogram-homology" (after test) are different
things and as such should be differently named.

>>(1) the parsimony itself (as already stated by others) does not
>>necessarily invoke the use of character congruence criterion in the
>>cladistic sense.
>I think it does. I agree that parsimony is a general principle, but its
>specific application is given meaning by the context of the analysis. The
>context for our analysis is that we have assembled a set of grouping
>hypotheses (characters), and we have a finite set of overall-relationship
>schemes (trees). ...
If you set the cladistic methodology as the "context of analysis", then the
congruence of course is the natural implementation of parsimony. Even within
the cladistics, however, you can think about alternatives - e.g. clique
analysis (=maximum subgraph analysis) *is* andother implementation of the
parsimony and is still cladistic in a sense. If you, on the other hand,
think about phylogenetic-tree-topology-approximation as the "context of
analysis", then you have many different parsimonious methodologies in hand.
Think about parsimony as what it really is: a minimization criterion placed
on some function (in the cladistics it is minimizing the number of character
changes); different clustering algorithms give you a flavor how different
such minimized functions can be.

>>(2) to "efficiently summarize the evidence" means to summarize under a
>>particular evolutionary model (also stated by others already).
>No. Why do you think this necessarily must be true?
Simply - different (assumed) evolutionary models *can* lead to different
character distribution along the phylogenetic tree and, consequently, also
within the extant species (or OTUs, individuals...). Then, different
methodologies are unequally successfull in approximation of the phylogenetic
tree topology - this success I call "efficiency". If, for example, rate of
change differ among characters, then the clique analysis (i.e. selection of
the globally congruent "informative" characters with the rejection of the
others) can be more successfull than classic cladistic analysis based on the
total character matrix.

>>Cladogram can summarize the evidence (in the sense of phylogen. tree
>>topology approximation) if and only if its construction principles apply
>> also to the phylogeny
>The phylogeny was "constructed" by the divergence of taxa.
I've meant that "construction principles of cladogram apply ...  to
phylogeny", of course.
>We assume that
>characters are passed from ancestors to descendants within these diverging
>lineages. We reconstruct the phylogeny by identifying the characters that
>are shared by the descendants.
As shown above, this can be done by different means under different
assumptions about evolutionary process.

>>Imagine, e.g., the evolutionary model in which each character would change
>>each 10th generation. Do yo mean that cladogram "efficiently summarize the
>>evidence" even under such conditions?

>The cladogram efficiently summarizes the evidence irrespective of any
model.
It does not - as shown above.

>I dont understand your question.
Apparently - this is probably the problem.

> Are you asking whether the cladistic summary would be
>efficient if life really had evolved as your model indicates? The answer is
>yes, - the cladogram is still the most efficient summary of the evidence.
>Whether it is correct is another matter.
Perhaps here is the point - what do you mean by "efficiency" if the output
is assumed to be "incorrect"? I believed that "efficiency" stands for "good
approximation to the phylogenetic tree topology" - what is yours definition?

>>moreover the underlying hypotheses about the character change pattern
>>... are basically untestable hypotheses (the "picture of evolution"
>>depends on the methodology which in turn depends on our assumed "picture
>>of evolution").
>Yes, this is problem of basing phylogenies on evolutionary models, and why
>we dont do that.
Some evolutionary model is always incorporated into analysis; of course,
some people do not know about it but this does not make the results better.
Best!
Zdenek Skala



---
Odchozí zpráva neobsahuje viry.
Zkontrolováno antivirovým systémem AVG (http://www.grisoft.cz).
Verze: 6.0.231 / Virová báze: 112 - datum vydání: 12.2.2001




More information about the Taxacom mailing list