Friday, January 28, 2011

Do they even understand the matrix they're coding?

You may remember my prior criticism of Hone and Benton (2008), who merged matrices when examining pterosaur relationships, but didn't combine equivalent OTUs.  They had both Lepidosauromorpha AND the lepidosauromorphs Gephyrosaurus, Sphenodontia and Squamata.  And they had both Choristodera AND the choristoderes Lazarussuchus, Cteniogenys and Champsosaurus.  Disturbingly, each more inclusive OTU did not clade with its exemplars in their trees.

Well now we have another example of using matrices without taking equivalent OTUs into account.  Yates (2004) had a sauropodomorph matrix where he tried two sets of outgroups.  One was Herrerasaurus and (Neo)Theropoda, another was Ornithischia and an expanded Theropoda including Herrerasaurus.  In the recent description of Chuxiongosaurus, Lu et al. (2010) used Yates' matrix but just ran the whole thing without noticing this.  Thus we have a cladogram with "Neotheropoda" sister to "Theropods"... :|  Herrerasaurus is the outgroup, probably because it is listed first in Yates' matrix, and Lu et al. didn't realize PAUP makes the first listed OTU the outgroup automatically.  Thus Ornithischia (which really should be the outgroup) is sister to the double-theropod clade.  What makes it especially sad is even if Lu et al. didn't read Yates' text, Yates labels the OTUs "Herrerasaurus ischigualastensis", "Theropoda (excluding Herrerasaurus)" and "Theropoda (including Herrerasaurus)" in his matrix.  How can you copy codings from that and not realize those aren't all independant taxa?!

As an aside, I'd like to see Chuxiongosaurus compared to Yunnanosaurus and Jingshanosaurus.  Thanks to Brad McFeeters for alerting me to this mess.

Coming soon, more Zhao taxa and part 2 of my recommended theropod clade definitions.

References- Hone and Benton, 2008. Contrasting supertree and total-evidence methods: the origin of the pterosaurs. Zitteliana. 28, 35-60. 
Lu, Kobayashi, Li and Zhong, 2010. A new basal sauropod dinosaur from the Lufeng Basin, Yunnan Province, Southwestern China. Acta Geologica Sinica. 84(6), 1336-1342. 
Yates, 2004. Anchisaurus polyzelus (Hitchcock): The smallest known sauropod dinosaur and the evolution of gigantism among sauropodomorph dinosaurs. Postilla. 230, 58 pp.

6 comments:

  1. "Yates (1994) had a sauropodomorph matrix where he tried two sets of outgroups."

    I think you meant Yates (2004) lol. :(

    ReplyDelete
  2. Hi, just a new reader here! Just wondering, how much of a 'amateur' paleontologist you are. Have you published papers or anything? Would be interesting to read your papers refutting all these other papers that your pointing errors out to. That's the way it's done, right?

    ReplyDelete
  3. Yes, Yates (2004)... *cough cough*

    As for writing papers which refute other papers, that's how it's traditionally done, yes. I'm one of the growing ranks of scientists who think that criticizing papers in blogs and other public media is also acceptable however. Ideally, the authors would see my post and not commit the same errors next time, thus solving the problem without having to go through the publishing process. Though in this case, I am actually submitting a manuscript with Marjanovic covering some of the coding issues I've raised on the blog.

    ReplyDelete
  4. Oh cool! What I rather meant actually was will it have an updated, corrected version of the data. Rather then just the mistakes made and not taken into account in the paper talked about here. Cos obviously that'd be more the interesting part to read, rather then the actuall refutting. Does that make sense? Hoping you'll let us know whenever it's out, like I already said, would be interesting to see :-)

    Wouldnt an easier way be to get into contact with the authors and tell them that way? Rather than putting it up on a blog and hoping that they'll see it? Or do you email them a link? Cos if they don't see it, then they don't learn their mistakes. And how does science move forward then? Don't papers quote other papers? They can't quote blogs though. So if a future paper wanted to reference the paper with some misinformation, then there's be no 'official' paper with a correction - how do they cite it then? Sorry for all the qsts - I'm new to all this and am trying to learn how the system works. Thanks for your time in answering!

    ReplyDelete
  5. It will have the updated version of some data, yes. In the present case, it would be easy enough to run Yates' matrix without the Neotheropoda+Herrerasaurus OTU and add Choxiongosaurus, but I honestly don't think the changes would be very interesting. Sauropodomorph relationships would probably be the same, and the matrix wasn't designed to determine basal dinosaur interrelationships, so any surprising result there could be chalked up to that.

    I actually find the refuting more interesting than the corrected results, because I think people expect trees to change given different data, but are quite ignorant of how terribly done many professional analyses are.

    Regardless of the difficulty in citing blog posts (which don't seem any worse than the oft used "pers. comm." in my opinion), there are simply too many papers with problems to consistantly publish papers on. The issue is better suited to blogs or features like the Evaluating Phylogenetic Analyses portion of my website. I know I'd be a regular subscriber to any Errors in Published Analyses Journal that dealt with such things, but I doubt that will ever appear.

    As for getting into contact with the authors, that's an interesting social issue. On the one hand, I think these problems definitely need to be made public since they affect papers people might otherwise use thinking they feature quality phylogenies. So a private email is out of the question. In addition to simply being more fun to write, I think my often scathing tone is deserved. These are professionals after all, who actually have the knowledge and capability to get this right. So when they don't code based on specimens I know they've seen and just copy from a prior matrix or (worse) don't code the state at all, I think they deserve to be called out on failing all of us who trust the raw data in their papers to be accurate. And I'm not afraid to do it under my real name, unlike some of my detractors here. But even given this, I think an email saying "go to this link to see where I ripped your paper a new one in public" would be a bit rude. Despite my issues with their work, I have no personal issues with most of these people. While I don't expect most of the authors to read my blog, I've found people usually get notified when they're being talked about online. So this is my compromise solution. But maybe it'd be a good idea to poll professionals to get their opinion...

    ReplyDelete
  6. Mickey

    I have a doubt about the actual size of Rajasaurus.
    Gregory Paul and other means we estimate about 11 feet and 4 tons. But the bones are too small to measure both. I know there are other examples as well as describe here, but I would like to compare with the holotype and if you could not help passing the measures.

    I show the quote that you made ...

    Holotype-(GSI No. 21141/1-33) (7-9 m) braincase, cervical centrum, partial dorsal vertebrae, sacrum, partial caudal vertebrae, partial scapula, partial ilia, proximal pubis, femoral (one distal), distal tibia , proximal fibula, metatarsals II, metatarsal IV 


    Referred-(ParaType of Lametasaurus indicus; lost) sacrum, ilium, tibia (in cranial reconstruction online) premaxilla, maxilla, lacrimal, jugal, postorbital, quadratojugal, dentary, teeth

    Thank

    ReplyDelete