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

RE: Clarification of scope of paleoart->uses

  While I would agree with the Wilson-based protocol of including appendices of 
such data, including them in the matrix itself is useless, as this information 
adds nothing to the analysis. 

  A. Problems with parsimony-uninformative characters.

  1. There is _no_ value for adding in characters which have one of two states 
that correspond only to one taxon. I would be dismissive of characters that 
exclude a basal form from all of the rest of the included matrix for this 
reason as well, as this optimizes the character as an autapomorphy of the 
outgroup as well as a synapomorphy of the ingroup. Mike Taylor notes 
(correctly) that such a character doesn't help you figure out a phylogeny, and 
this is precisely why it shouldn't be included in a phylogenetic analysis. The 
value of the autapomorphy to the matrix is _nil_. Its value outside of the 
analysis is more useful, as in the case of a series of autapomorphies, and this 
can be done without ever resorting to a phylogenetic analysis.

  2. Parsimony-uninformative characters increase processing time (not enough to 
be significant unless they are in large numbers). 

  B. Solutions to matrix construction.

  3. A matrix database of characters (such as what Paul Sereno is attempting) 
adds to the information of characters one may include: A drop-and-drag feature 
from such a database allows the analyst to grab and code (unless it also comes 
with the appropriate coding per taxon) any character at his disposal which is 
parsimony informative, leaving out what is essentially empty noise.

  4. An analyst may include absolutely every single character he can think of 
and plug it into a matrix of the appropriate OTUs. This can be done for various 
reasons, including

  a. Running a distribution-pattern analysis, without invoking the illusion of 
phylogeny: this is done with language-based cladistic analysis. 

  b. It can also be useful for assessing patterns of expression in shape under 
an objective heading. Mark Witton recently responded to my comment about the 
TMM snout being a tapejarid, whereas I had forgotten his assessment in 
discriminating chaoyanopterids, tapejarids, "tupuxuarids" and thalassodromids 
within *Neoazhdarchia*.

  c. Such an analysis can help by assessing the value of size, shape, and 
proportion-based characters which are also typically excluded from phylogenetic 
analysis, can be used to discriminate age-classes among taxa without resort to 
morphospace analysis (which nonetheless functions similarly).

  I think Mike is looking at the value of the matrix beyond the recourse to 
phylogeny, so this response is not a retort but a supportive frame of 
reference. The problem is that one should not pretend that the analysis is 
phylogenetic when doing so, and separating characters beforehand can be useful 
when one knows they lack the distribution to be informative to phylogeny. 
Removing them after the fact when one finds that they aren't informative 
improves the matrix. I understand there is a potential feeling that having the 
bigger dataset may improve the information garnered, but when that info is 
partly junk, and the analyst both knows and retains it, then he is not doing 
Science service.  


Jaime A. Headden
The Bite Stuff (site v2)

"Innocent, unbiased observation is a myth." --- P.B. Medawar (1969)

"Ever since man first left his cave and met a stranger with a
different language and a new way of looking at things, the human race
has had a dream: to kill him, so we don't have to learn his language or
his new way of looking at things." --- Zapp Brannigan (Beast With a Billion 

> Date: Wed, 16 Mar 2011 22:49:11 +0000
> From: mike@indexdata.com
> To: david.marjanovic@gmx.at
> CC: dinosaur@usc.edu
> Subject: Re: Clarification of scope of paleoart->uses
> On 16 March 2011 22:37, David Marjanovic  wrote:
> >>  For my cladistic analysis, i lay out everyone else's matrix and then
> >>  toss those characters that, somehow, i determine to be most
> >>  unrelable.
> >
> > Don't.
> >
> > There's no reason to throw out any characters that aren't
> > parsimony-uninformative (and even those don't actually hurt);
> I'd go further. I think it's important to retain
> parsimony-uninformative characters for two reasons.
> The obvious one is that, if someone goes on to build on your matrix,
> the new taxa they add may make the previously uninformative characters
> informative.
> The less obvious reason is that one of the ways a character can be
> parsimony-uninformative is if it occurs in only a single taxon. In
> this case, it doesn't help you figure out the phylogeny, but it DOES
> tell you something that's distinctive about the taxon in question
> compared with the next largest clade that it's a member of. It's part
> of the differential diagnosis of that taxon.
> One of the reasons that Wilson (2002) is still in many respects a more
> useful sauropod phylogeny that than other that have appeared since,
> some with far larger matrices, is that it has whole delicious
> appendix, on pages 271-276, that lists autapomorphies for all the taxa
> in the analysis. [To be fair, Upchurch et al. (2004) also did this,
> but editorial style imposed the unhelpful decision that the
> autoapomorphies had to be sprinkled through the text and described in
> prose rather than telegraphically.] The autapomorphies in Wilson's
> list are of two kinds: those that are homoplastic in the phylogenetic
> analysis, and which are therefore autapomorphic only in specific
> cladograms; and others that were not included in the analysis. The
> appendix's introduction doesn't spell it out, but I am guess that many
> if not all of these were discovered as parsimony-uninformative
> characters.
> So: don't throw data away.