Nocturnal dinosaurs, again: the benefits of discriminant analysis in quantitative paleobiology

More on nocturnal dinosaurs, or, importantly, on our quantitative approach to infer ecology and behavior in fossil vertebrates. Hall et al.’s technical comment provided us with the opportunity to further explain the benefits of our approach.

Schmitz, L. & R. Motani (2011). Response to Comment on ”Nocturnality in dinosaurs inferred from scleral ring and orbit morphology”. Science, 334: 1461. html, reprint.

Quantitative paleobiology is on the rise, with more and more researchers using structure-function relations to infer the ecology and behavior of long extinct organisms. A new and powerful method is phylogenetic Flexible Discriminant Analysis (pFDA). pFDA provides a probabilistic framework for paleobiological analyses, resulting in testable, quantitative models of ecomorphological inferences.

I don’t want to spend too much time dissecting Hall et al.’s comment (Ryosuke and I did this in detail in our response). Most importantly, the inference of diel activity patterns in Mesozoic archosaurs is sound. However, in more general terms, I would like to stress the importance of a few key points of the ecomorphological inference in fossil vertebrates in general and pFDA specfically:

(1) It is very important to have a good understanding of the structure-function relations of the functional system in question. In other words, one needs to understand the functional implications of a morphological structure. This sounds trivial but it’s not.

(2) It is absolutely necessary to incorporate phylogenetic information. Even a supposedly weak phylogenetic signal may have profound effects, as has been shown in many studies. Luckily, well supported and time-calibrated phylogenies are becoming available for more and more clades.

(3) Whenever possible, prior probabilities for pFDA should be used, because they contain very important and biologically meaningful information. If prior probabilities are unspecified, discriminant analysis attempts to group samples into equal proportions, which (in most cases) is unwarranted.

I will expand the discussion of each of the points listed above in the (hopefully) near future. Also stay tuned for a pFDA guide in R. Such a step-by-step guide is already available in the SOM of the original paper, but it seems necessary to make it easier to access. Until then, please let me know if you have questions regarding pFDA.

This entry was posted in fossils, own research, vision and tagged , , , , , , , , , , , . Bookmark the permalink.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s