The interest in the results of a PCA are usually not how much variation can be accounted for with each axis. The interest is in the ordination plot that gives you a low-dimensional approximation of the high-dimensional shape space. You can then appreciate whether the data look like a single random cloud of points or if there is interesting structure that might lead you to biological insights.
========================= F. James Rohlf Distinguished Professor, Stony Brook University http://life.bio.sunysb.edu/ee/rohlf > -----Original Message----- > From: morphmet [mailto:[EMAIL PROTECTED] > Sent: Wednesday, October 10, 2007 9:08 PM > To: morphmet > Subject: Morphometrics overkill for this problem? > > Thanks Colin and Goran for your replies. Colin, we've actually got > several thousand optic disc photos which I plan to do a trace of, > but > thanks for the information anyway. > > One other question; > > I've read that you can break down Fourier co-efficients using > principal > component analysis (and have actually been having a play with the > 'Shape' software available on this site). I understand that PCA is > a > data reduction technique. However, I don't come from a maths > background > and I'm having trouble visualising what PCA actually gives you in > the > break-down analysis of a shape. For example, using this software, > for a > sample of outlines > > Eigenvalue Proportion(%) Cumulative(%) > 1/96 > > Prin1 3.554525E-004 53.9578 53.9578 * > > Prin2 2.099497E-004 31.8704 85.8282 * > > Prin3 6.445964E-005 9.7850 95.6132 * > > Can someone explain (in lay terms) what the numbers actually mean > (in my > situation of an optic disc)? Are they vectors describing direction > and > magnitude? I can't seem to find any good resources online that > explain > PCA without a considerable background maths knowledge basis (maybe > there > is no simple way!) > > Is PCA an accepted scientific data analysis technique? > > Thank you. > > Paul > > [EMAIL PROTECTED] > > -- > Replies will be sent to the list. > For more information visit http://www.morphometrics.org -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
