[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
RE: Erketu ellisoni online free access
Ken Carpenter wrote-
Ah, not quite. Appendix 2 gives the coding for Erketu from the published
matrix of Wilson 2002. >Of 234 characters only 18! can be coded for Erketu.
That is 7.7% of possible characters! How can >anyone tell me that we can
assume to know so much from so little???? Keep in mind that a >phylogenetic
analysis is simply a form of statistics. Even Wiens' article notes a
decrease in reliability >of phylogenetic results when 75% or more of the
data is unknown (his Fig 2). Here 92% is >unknown!!!!
In case you were referring to figure 2 of Wiens' 2003 paper, it actually
means something else. It shows that adding a set of 50 new characters to a
50 character analysis is beneficial unless those new characters are unknown
in 75% or more of the taxa. And even then, they're not harmful to add. So
it's about adding incomplete characters to analyses (like integumentary or
eggshell characters in Mesozoic dinosaur analyses), not incomplete taxa.
What you want to refer to is figure 1, where the affect of incomplete taxa
(like Erketu) is analyzed. Wiens notes- "Given enough characters in the
analysis, it is possible to have extremely accurate resolution when
including taxa that are only 5% complete and that have nearly 2000 missing
data cells each (Fig. 1). Clearly, the amount of missing data itself is not
the actual problem. As long as enough characters are sampled in the
incomplete taxa to accurately place them on the tree, then the amount of
missing data seems to have little impact."
Wiens' later (2006) paper has equivalent graphs to figure 1 for a 200
character analysis. In particular, figure 2b of Wiens (2006) is relevant*,
being a parsimony analysis where the characters taxa lack data for vary
among those taxa (as in sauropod analyses where some taxa are only known
from skulls, some lack skulls entirely, etc.). It shows that even when
_half_ the taxa have 90% of their character states unknown, the results are
still over 50% accurate. Of course, every other taxon in Wilson's analysis
is more complete than Erketu, and the average amount of unknown data is 45%,
even taking Erketu into account. So the accuracy should be >90%, according
to Wiens' graph.
Wiens' analyses do show that we need to use more characters in our
morphological analyses however. Are 2000 characters too much to ask for in
a sauropod analysis? I don't think so. And if we can reach that point, we
can have 100% accuracy even if our matrix is 75% incomplete.
* So if you were referring to figure 2 of Wiens 2006, and not Wiens 2003,,
ignore my first paragraph.