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

Re: D.Marjanovics repliesto J'r pts I-II



> >         There are a few papers whose authors have looked at every
specimen
> > they've coded. HP Oliver Rauhut's dissertation and follow-up paper are
> > examples.
>
> Not true.  Rauhut didn't personally examine Euparkeria, Staurikosaurus,
> Herrerasaurus, "Szechuanoraptor", Poekilopleuron, Torvosaurus,
> Monolophosaurus, Siamotyrannus, Sinraptoridae or Bagaraatan.

Oh. OK. Well, one more example of those where most were seen in person. I
should also mention that most of the animals in his matrix are better
described than most Chinese EK birds.
        I wanted to add another point in my post, but forgot: Based on the
description, I could most probably code *Jixiangornis* for a few more
characters than I meanwhile have (not online yet). But I can't _read_ the
description. It's in Chinese. I can just figure out which paragraph is about
which bone. I have to be content with the quite simple line drawings and the
rather small and dark photos (the fact that I got the paper as a scan
certainly doesn't help).

> Peters is correct.  It's simply not feasible to see every specimen, and
> people do write articles in the literature for a reason.

That's usually right. Still, my analysis clearly stays unpublishable -- I
could double the number of coded characters for many OTUs in my matrix if I
would investigate a specimen, so there are a large number of unnecessary
question marks in my matrix which decrease the resolution of the trees and
probably add some artifacts.
        I certainly don't need to see *Ichthyornis*, for example. It now has
a gorgeous description. But *Eoenantiornis*...

> And even if you code off someone else's matrix, your data is
> at least as correct as theirs, and their
> article was apparently fit to publish.

What was fit to publish 10 or just 5 years ago is often not fit to publish
today -- for good reasons.

> I just think it's ridiculously idealistic to think
> data isn't publishable until all specimens are personally examined.

I implied something too close to this than I wanted...

> >         In short, we _are_ unable to finish the job in the foreseeable
> > future. Period.
>
> I think Peters has a point here too.  With the small amount of characters
> and taxa in nearly any dinosaur or pterosaur cladogram, the authors
> certainly haven't tried hard to "finish the job".  Holtz (2000) is the one
> near exception among coelurosaur cladograms.  My analysis may have >400
> characters, but I haven't even begun to try "finishing the job".  It may
not
> be possible to get a confidently true tree from our small sample sizes,
but
> it's usually possible to get a well resolved tree, ignoring instances
where
> related taxa preserve no elements in common.

Yes. But we shouldn't -- full circle -- regard every analysis as
unpublishable or otherwise inferior that has more than one most
parsimonious/likely tree. Having several OTUs with much missing data makes
it almost automatically impossible to find only one best tree; still, you
have demonstrated why they should be included, and so have the theoretical
studies in JVP 23(2). To find out where the equally best trees agree and
where they don't often yields valuable insights.