[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
RE: Dollodon status questioned
Liz -
Actually, running a PCA on three variables is a perfectly reasonable thing to
do, especially if you want to, say, extract a size axis/variable based on those
three variables and if you reasonably consider PCA more of a mathematical
method than a statistical one. You can even do it on two variables, which then
essentially is the basis for doing major axis line fitting (as opposed to
reduced major axis or standard regression). So no statisticians should infarct
over it.
That the PCA will give you, most probably, a PCI that is a huge size factor,
especially for dinosaurs, especially for sauropods, is indeed true as well
depending on the size distribution of the animals measured. There are some
animals, however, where size does include taxonomically important data - bats,
for example, but they tend to have an adult size rather than grow continuously
and it is especially common when adult size is really constrained by, say,
flying. Also, sometimes the taxonomically important variation gets caught up in
that size data and it gets tricky to extract/interpret. So one would want to be
retrospective about using it as such for sauropods but, depending on the sample
size and size distribution, there could be some neat info there.
R
-----Original Message-----
From: owner-DINOSAUR@usc.edu [mailto:owner-DINOSAUR@usc.edu] On Behalf Of
Denver Fowler
Sent: Friday, March 25, 2011 12:32 PM
To: DML
Subject: Re: Dollodon status questioned
Forwarded for Liz Freedman
On Fri, Mar 25, 2011 at 1:36 AM, Jaime Headden <qi_leong@hotmail.com> wrote:
> I wonder if McDonald uses some more substantive reasoning (optimistically,
> I
>assume he does) than in the abstract -- which is, let's face it, an abstract.
>I
>haven't the paper, but will soon.
>
Not really - he runs a PCA on a mere 3 measurements for each specimen
(which would give any statistician a heart attack), and those 3
measurements are all known to vary with the size of the specimen
(i.e., likely ontogenetically). Axis 1 (80% of variance) is probably
based entirely on the size of the specimen, with no real taxonomic
value.
-Liz Freedman