While there is a certain elegance to computing partial warps and the uniform component separately, it is a mathematical elegance not a biological one. In practice they should almost always be treated together to provide a set of shape variables that span the tangent space that is used for multivariate analyses of shape variation.
There are two strategies that can be used. One can either compute the partial warps and the uniform component and then project the GPA aligned specimens onto them or one can simply do a PCA of the GPA aligned specimens if the partial warps and uniform components are not of interest in themselves. That saves a computational step. In both cases one has to be careful of how the results of the GPA are scaled. The literature is not entirely consistent (full vs. partial Procrustes fitting). When variation in shape is very small the distinction will make little difference. For general software one must be prepared for a user to try it on a case where variation is much larger. I, of course, like the way I do it in the tps software. ---------------------------------------------- F. James Rohlf, Distinguished Professor Dept. Ecology & Evolution, Stony Brook University Stony Brook, NY 11794-5245 http://life.bio.sunysb.edu/ee/rohlf > -----Original Message----- > From: morphmet [mailto:[EMAIL PROTECTED] > Sent: Friday, February 16, 2007 5:46 PM > To: morphmet > Subject: Affine component in relative warps > > Hi all, > > For computing relative warps in my program PAST, I have used > the approach given by in the book Statistical Shape Analysis > by Dryden & Mardia. The formulations differ a bit from those > given by Bookstein, Rohlf etc. > but are (hopefully!) equivalent. > > Now the problem is that I feel obliged to include the affine > component, which is not described by Dryden & Mardia - they > just calculate it separately and do not integrate it with the > other relative warp components. > > I admit I am confused by the methods for including the affine > component as used e.g. in tpsRelw, and even more mystified as > to how to do a similar thing within the framework of Dryden & Mardia. > > Has anyone, by any chance, run into a similar problem and > solved it? I would be grateful for any advice! > > (I suppose I could start the programming all over and follow > e.g. Rohlf instead of Dryden & Mardia, but this would be a > little work ...) > > > Thanks, and regards, > > Oyvind Hammer > Natural History Museum > University of Oslo > > > > -- > 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
