Asta, I appreciate the complexity of your problem and would like to make an initial comment that might direct your reading on testing the usefulness of parallel data. If you have metric data of different types from the same individuals you might benefit from looking at Rao's Analysis of Dispersion model in which you can have multiple fields of Y data for each record and then test for whether there is any additional information in an inclusive set of traits given a subset. This would seem to be an important part of testing and possibly simplifying your design.
Rao, C. R. (1965) Linear Statistical Inference and Its Applications. John Wiley & Sons, New York, 522pp. Joe Kunkel On Tuesday, February 4, 2003, at 10:43 AM, [EMAIL PROTECTED] wrote: > Hello, > > for some time I am trying to figure out the most proper design for my > study and greatly appreciate any comments. > > I want to study intraspecific geographic morphological variation in > mysid shrimps (Mysis relicta) and correlate/compare it with the > presumably neutral molecular variation (estimated from allozymes and > mitochondrial DNA gene sequence) and certain ecological variables (e.g. > latitude, depth, size, nutrient conditions of the lake, its age and > drainage basin). I plan to include about 10-15 populations distributed > in N. American lakes, and for the beginning use about 10 specimens from > one population. > >> From the pilot studies I expect the morphological differences among > populations to be comparatively small. In order to get the best > estimate > of overall differentiation and analyse response of different > shapes/characters I intend to evaluate about 50 characters, both linear > measurements from articulated structues and data from landmarks on 2 or > 3 structures (6-9 landmarks on each). > > I have two questions related to this study: > > 1. What is the best way to analyse landmark data, if I want to compare > within to between population variance and to get ordination of > populations (I expect that variation will be small). What is a > difference of using multivariate statistical analysis on the residuals > after Procrustes superimposion or on the matrix of partial warp > scores+uniform component (in addition to statistical power - Rohlf > 2000)? > > 2. I first plan to compare differentiation of different shapes and > linear measurements separately, but -- is there any proper way to > combine data from landmarks and linear measurements to obtain overall > measure of differentiation among populations? For example, would it be > possible to combine matrices of partial warp scores from different > structures and presumably "size-free" linear measurements (e.g. > residuals of raw data to regression on PC1) into one matrix > and use it for PCA or other analysis? > > many thanks in advance > Asta > > ** > Asta Audzijonyte (PhD student) > Department of Ecology and Systematics & > Finnish Musuem of Natural History > University of Helsinki, Finland > [EMAIL PROTECTED] > == > Replies will be sent to list. > For more information see > http://life.bio.sunysb.edu/morph/morphmet.html. > > ----------------------------------------- Joseph G. Kunkel, Professor Biology Department University of Massachusetts Amherst Amherst MA 01003 http://www.bio.umass.edu/biology/kunkel/ == Replies will be sent to list. For more information see http://life.bio.sunysb.edu/morph/morphmet.html.
