-------- Original Message -------- Subject: Re: Comparing variability Date: Mon, 23 Jun 2008 06:24:32 -0700 (PDT) From: andrea cardini <[EMAIL PROTECTED]> To: morphmet@morphometrics.org Dear Louis, this seems to be similar to something asked recently by another subscriber (on June 2nd, if I am correct). Please, find below my answer to that message. There's more in the MORPHMET archive. The way you suggest to compute shape variances looks the same as Philipp's suggestion: "the trace of the covariance matrix (sometimes called total variance), which equals the sum of all univariate variances or the sum of all eigenvalues". All the best Andrea if you want to see whether variance is larger within a population compared to another one, as one would do for instance to test for a reduction in variance in an island population, this is something that you can possibly do with a Levene's test. An example is described in a paper of mine (below) and another one (or more), if I remember well, can be found in studies by Hallgrimsson and colleagues. To get the Procrustes distances to the mean you can simply use TPSSmall for both 2D and 3D data. Tests can then be done in any statistical software but to have also permutations it might take a little more effort unless you're using NTSYS. I am pretty sure that most of these tests (certainly MANOVA/DA but possibly also tests for variances - check the disparity module) are also available in the IMP series, where you generally have options for doing resampling stats too. Good luck Andrea Cardini A, Thorington Jr. R. W., P. D. Polly, 2007 - Evolutionary acceleration in the most endangered mammal of Canada: phylogenetic signal and cranial divergence in the Vancouver Island marmot (Rodentia, Sciuridae). Journal of Evolutionary Biology, 20: 1833-1846. "The LeveneÂs test requires calculating the absolute value of the deviation of each individual from the sample mean (Van Valen, 1978), which is satisfied by using absolute differences to the mean size or Procrustes distances to the mean shape. These deviations are then compared by ANOVA ANOVA. Although this test is generally considered relatively robust to departures from normality, we chose to perform a randomization version of the test that compares the observed F-statistic with the distribution obtained by randomly reassigning deviations from sample means to the samples." At 08:33 23/06/2008 -0400, you wrote:
-------- Original Message -------- Subject: Comparing variability Date: Mon, 23 Jun 2008 02:13:43 -0700 (PDT) From: Louis Boell <[EMAIL PROTECTED]> To: <morphmet@morphometrics.org> Dear colleagues, I want to compare variabilities of Procrustes coordinates across groups. Is it adequate to simply calculate the variance of each coordinate for each group, to sum the values over all coordinates and to use the result as a measure of shape variability? Best regards, Louis Louis Boell MPI für Evolutionsbiologie August-Thienemannstr.2 24306 Plön [EMAIL PROTECTED] [EMAIL PROTECTED] ------------------------------------------------------------------------ Zeigen Sie es Ihren Freunden! Erstellen Sie kostenlos Fotoalben und Diashows in Ihrem Space! <http://redirect.gimas.net/?cat=hmtl&n=M0804Fotoalbum&d=http://spaces.live.
com/api.aspx?wx_action=addModule&mkt=de-DE&wx_ru=http%3a%2f%2f%7bspacesapi%3 aid%7d.spaces.live.com%2fphotos%2f&wx_targetSite=PersonalSpaces&wxp_url=%2fp hotoalbum.aspx>
-- Replies will be sent to the list. For more information visit http://www.morphometrics.org
Dr. Andrea Cardini Lecturer in Animal Biology Museo di Paleobiologia e dell'Orto Botanico, Universitá di Modena e Reggio Emilia via Università 4, 41100, Modena, Italy tel: 0039 059 2056532; fax: 0039 059 2056535 Honorary Fellow Functional Morphology and Evolution Unit, Hull York Medical School University of Hull, Cottingham Road, Hull, HU6 7RX, UK University of York, Heslington, York YO10 5DD, UK E-mail address: [EMAIL PROTECTED], [EMAIL PROTECTED], [EMAIL PROTECTED] http://hyms.fme.googlepages.com/drandreacardini http://ads.ahds.ac.uk/catalogue/archive/cerco_lt_2007/overview.cfm#metadata -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
