[Taxacom] sloppy cladistic analyses
phovenkamp at casema.nl
phovenkamp at casema.nl
Thu Feb 4 07:57:56 CST 2010
BTDT (Been There Done That)
For a brief and fairly shallow exploration of the severity of coding errors see my paper from 1999 (Unambiguous Data or Unambiguous Results? Cladistics 15: 99-102). Since then, I've occasionally amused myself with applying the same permutation method to other analyses (morphological ones, mostly), generally finding that there is usually a number of clades that turn out to be unstable. But what level of unstability would be acceptable is up for debate (in a reasonably sized dataset of say, 20 taxa 50 characters, a stability of 99 % equates to c. 10 matrix entries the value of which is essential for the "survival" of the clade.
Best,
Peter Hovenkamp
----- Origineel bericht -----
Van: Stephen Thorpe <s.thorpe at auckland.ac.nz>
Datum: woensdag, februari 3, 2010 21:23
Onderwerp: Re: [Taxacom] sloppy cladistic analyses
Aan: John Grehan <jgrehan at sciencebuff.org>, "taxacom at mailman.nhm.ku.edu" <taxacom at mailman.nhm.ku.edu>
> > The problems outlined here by Thorpe (coding errors, mising data,
> subjectivity, polarity estimations, outgroup choice) are not just
> those of cladistics, but of systematics in general
>
> Perhaps Grehan is correct, but I still think that the problems are
> worse for cladistics, or at least any methodology in systematics which
> depends heavily on "number crunching". They have a tendency to lose
> the "narrative" - a huge data matrix and associated tree are not a
> narrative! Each and every step to a conclusion needs to be fleshed out
> and critically evaluated. The fleshing out is important in order to be
> able to spot errors which are otherwise virtually impossible to see
> within a huge data matrix. I'm not sure if there have been any studies
> done on the amount of influence on the conclusion that various amounts
> of coding errors can have? "People" have said to me "that doesn't
> matter!" when I have pointed out coding errors, but how do they know
> it doesn't matter? Matter to whom? For every coding error I spot, how
> many others are there? At what level does it start to matter?
> ________________________________________
> From: taxacom-bounces at mailman.nhm.ku.edu
> [taxacom-bounces at mailman.nhm.ku.edu] On Behalf Of John Grehan [jgrehan at sciencebuff.org]
> Sent: Thursday, 4 February 2010 2:06 a.m.
> To: taxacom at mailman.nhm.ku.edu
> Subject: Re: [Taxacom] sloppy cladistic analyses
>
> The problems outlined here by Thorpe (coding errors, mising data,
> subjectivity, polarity estimations, outgroup choice) are not just those
> of cladistics, but of systematics in general. Similarity, the problem
> identified by Ken with cladistic analyses where codings of previously
> analyses are cut and pasted with their own without critical evaluation.
> This is a problem that is not necessarily confined to cladistic analysis
> since any type of systematics may be underminded by this practice.
> Perhaps one of the few explicit attempts to demonstrate that probmem
> has
> been with the recent hominid analyses where previously claimed
> characters have been subject to detailed critique whereas other
> analyeses claiming to support the chimpanzee relationship are plagued
> with the cut and paste recycling method - which of course means that
> they always get the same answer more or less.
>
> I realize that Ken and various others are opposed to cladistics and that
> is fine with me, but to attribute general problems of systematics to
> cladistics alone is (do doubt unintentionally) misleading.
>
> John Grehan
>
>
>
> > -----Original Message-----
> > From: taxacom-bounces at mailman.nhm.ku.edu [mailto:taxacom-
> > bounces at mailman.nhm.ku.edu] On Behalf Of Stephen Thorpe
> > Sent: Tuesday, February 02, 2010 10:52 PM
> > To: Kenneth Kinman; taxacom at mailman.nhm.ku.edu
> > Subject: Re: [Taxacom] sloppy cladistic analyses
> >
> > I'm not sure that the internet is to blame here. It seems to me to
> be
> more
> > an intrinsic problem to cladistic analysis itself. Anything that
> involves
> > numbers is prone to transcription errors, and the nature of some
> people
> > increases the chances of this happening. How many cladistic analyses
> get
> > checked for coding errors? This problem on top of the other major
> problem
> > that most of the relevant data is missing (because only a minute
> fraction
> > of taxa have been informatively preserved as fossils), not to mention
> > subjectivity in character weighting, polarity estimations, and
> outgroup
> > choice, and what is the worth of such analyses???
> >
> > ________________________________________
> > From: taxacom-bounces at mailman.nhm.ku.edu [taxacom-
> > bounces at mailman.nhm.ku.edu] On Behalf Of Kenneth Kinman
> > [kennethkinman at webtv.net]
> > Sent: Wednesday, 3 February 2010 4:18 p.m.
> > To: taxacom at mailman.nhm.ku.edu
> > Subject: [Taxacom] sloppy cladistic analyses
> >
> > Dear All:
> > Although I don't ALWAYS agree with Michael Mortimer, his
> cladistic
> > analyses are far better than most. Therefore, I find his following
> > critique of many recent cladistic practices and shortcomings very
> > seriously. It reflects a broader problem among computer generated
> > so-called "information" and an alarming trend of internet
> > DISINFORMATION now competing with or even outpacing good information.
> > What one now finds on the internet, including scientific
> > information, must increasingly be taken with a huge grain of salt.
> The
> > truism about computers in the hands of more sloppy users is
> > unfortunately an increasing reality: "garbage in, garbage out." This
> > is certainly true of cladistic analyses by those who just don't
> > critically evaluate the codings of previously analyses and just cut
> and
> > paste them and add a few of their own. Adding a little new
> information
> > to a database riddled with garbage, and the garbage can overwhelm the
> > new information (whether the new information might be helpful or not).
> > As Mortimer says, it can give a false impression of consensus in
> > something that may or may not be true. Here's a link to his concerns:
> >
> > http://dml.cmnh.org/2010Feb/msg00010.html
> >
> >
> > _______________________________________________
> >
> > Taxacom Mailing List
> > Taxacom at mailman.nhm.ku.edu
> > http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
> >
> > The Taxacom archive going back to 1992 may be searched with either
> of
> > these methods:
> >
> > (1) http://taxacom.markmail.org
> >
> > Or (2) a Google search specified as:
> > site:mailman.nhm.ku.edu/pipermail/taxacom your search terms here
> > _______________________________________________
> >
> > Taxacom Mailing List
> > Taxacom at mailman.nhm.ku.edu
> > http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
> >
> > The Taxacom archive going back to 1992 may be searched with either
> of
> > these methods:
> >
> > (1) http://taxacom.markmail.org
> >
> > Or (2) a Google search specified as:
> > site:mailman.nhm.ku.edu/pipermail/taxacom your search terms here
>
> _______________________________________________
>
> Taxacom Mailing List
> Taxacom at mailman.nhm.ku.edu
> http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
>
> The Taxacom archive going back to 1992 may be searched with either of
> these methods:
>
> (1) http://taxacom.markmail.org
>
> Or (2) a Google search specified as:
> site:mailman.nhm.ku.edu/pipermail/taxacom your search terms here
> _______________________________________________
>
> Taxacom Mailing List
> Taxacom at mailman.nhm.ku.edu
> http://mailman.nhm.ku.edu/mailman/listinfo/taxacom
>
> The Taxacom archive going back to 1992 may be searched with either of
> these methods:
>
> (1) http://taxacom.markmail.org
>
> Or (2) a Google search specified as:
> site:mailman.nhm.ku.edu/pipermail/taxacom your search terms here
>
More information about the Taxacom
mailing list