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

Re: Wukongopterus and Darwinopterus



At last...

 David M. wrote with regard to phylogenetic analysis:

 > Of course, the solution isn't to exclude data, it's to include even
 > more -- and to recheck the _quality_ of the data: How many typos are
 > in there? How many correlated characters? Which characters should be
 > ordered or given a stepmatrix, and why? And so on and so forth
 > (Jenner 2001; Marjanovic & Laurin 2008b, 2009; Marjanovic, Germain &
 > Laurin in prep.). Size does matter (Emmerich 1998), but so does
 > quality (Kitamura 2004).

 When the tree topology is correct typo corrections will support the
 tree. When not, typos will correct a tree.

As Mike Habib has mentioned, this is not always how datasets behave in reality. Any changes to a data matrix have _unpredictable_ effects, unpredictable in both kind and magnitude. Changing a single cell can leave the topology intact and just make microscopic changes to support values and the length of one branch -- but it can just as easily overturn the entire topology. I'm sure I've recommended fig. 10 of Warren's paper in the December 2007 issue of JVP before.

I have also seen it happen several times during my work that corrections change a topology, and then further corrections change it back. The same, of course, can happen to support values -- when I started working on the matrix I'm now working on, the lepospondyl hypothesis on the origin of lissamphibians was 1 step more parsimonious than the temnospondyl hypothesis, then the difference it rose to 7 or 8 or something, then it came back down to 4, and then it rose back up to 9, where it is now, and where it might not stay.

 Correlated characters? How difficult to determine: 1) long neck vs.
 long tail on sauropods;

Let's see if all four possible combinations exist:
- Long neck, long tail: diplodocids.
- Short neck, short tail: ...are there any sauropods that count, actually...?
- Short neck, long tail: dicraeosaurids.
- Long neck, short tail: *Giraffatitan* sort of.

I'd say they're not correlated. Of course, for an actual published study as opposed to a casual e-mail, it would be a good idea to actually measure that stuff and calculate the correlation coefficient.

 2) short manual toes vs. short pedal toes on stegosaurs;

Certainly correlated -- a single adaptation for quadrupedal graviportal locomotion.

 3) wide skull vs. orbits on skull roof on batrachomorphs;

That term has fallen out of fashion. Do you mean temnospondyls and lissamphibians? When the skull is so flat that it has no lateral side, the orbits will be on the dorsal side, obviously; this means that, when the skull is flat (in which cases it will also be wide), the character that describes the position of the orbits must be scored as _inapplicable_. Taller skulls like a hippo's can have the orbits in several positions.

 4) number of caudal vertebrae versus loss of teeth in birds, etc.
 etc. etc. The list is endless.

So what? That it's time-consuming and difficult does _not_ mean we shouldn't even try. To the contrary -- if it were completely impossible, _phylogenetics would be completely impossible!!!_ Ignoring a problem never makes it go away.

 ordered vs. unordered: so difficult to determine with convergence and
 parallelism.

That doesn't even enter the question. Ordering is something that's determined before the analysis. The states of a potentially continuous or meristic character must be ordered (Wiens 2001, Syst. Biol.) because you already used that assumption for partitioning the character into states in the first place -- the assumption that it's easier to go from a state to a similar state than to a morphologically distant one.

Yes, there are cases where it's tricky, and there are cases where both ordering and not ordering are too simple and stepmatrices are needed. Cry me a river. Phylogenetics is hard work.

 Best to just let the suite have its say-so if there's any question.

If you have 3 characters in your matrix that are correlated to each other, that's the same as having 1 character in it that shouts 3 times as loud as the others. Because you gloss over this problem, the say-so you get is distorted.

 Evolution abhors apomorphies, preferring parsimony.

You're making this up.

You are making this up.

"...and stop anthropomorphizing her, she hates that." There is no mechanism that informs a mutation that it has already happened elsewhere and therefore must not happen again. What did you think?!?

 Also, a good, healthy tree of
 sufficient size can sustain many dozens of typos, if not all
 concentrated near a weak branch.

This seems logical, but in fact it's backwards. The denser the taxon sampling, the shorter every internode will be -- and the fewer changes to the matrix will be required to overturn any of them.

I had this happen in my work. In this paper http://dpc.uba.uva.nl/cgi/t/text/text-idx?c=ctz;sid=d42a482bca8fcb6eec9a4635192c370d;tpl=browse-toc-77.tpl (scroll down to number 3) I added a taxon (*Brachydectes*) that broke up a long internode, and the bootstrap values in the clade that it held (Albanerpetontidae + Lissamphibia) _sank_ (fig. 6). That's perhaps counterintuitive, but it's logical: the added taxon is similar enough to that clade that, in some bootstrap replicates, it entered it.

 "Weasel" words (i.e. could, can, might, may, would, should, possibly,
 recheck the quality, correlated, heterodox, believe, etc. ) are signs
 that specific data are being side-stepped. "An apple a day keeps the
 doctor away" is a falsifiable statement. An apple a day (can, may, is
 believed to) keep the doctor away" is not. I trust no referee who
 relies on weasel words when falsifiable data is on the table.

That's short-sighted of you. I used "can" because it happens in some cases (pers. obs.) but not in others (pers. obs.). It's not predictable.

Also, I'm allergic to apples.