-------- Original Message -------- Subject: Comparing shape variances Date: Fri, 23 Jan 2009 02:27:53 -0800 (PST) From: Louis Boell <[email protected]> To: <[email protected]> Dear colleagues, I would like to ask for your advice on the following problem: I want to compare shape variances of landmark/GM data of three different samples of mouse mandibles. I want to find out whether the variances are significantly different between samples. My problems are: 1) should I use the procrustes variances or rather the variances of some derived variables, e.g., PCA scores? 2) is it appropriate/possible to use univariate tests like F-Test, Levene´s test or Bartlett´s test in the special multivariate context of GM data? Can I assume normality in these data? 3) How should I deal with very unequal sample sizes (between 15 and 130)? 4) Knowing the pedigrees of the mice, would it be an alternative to calculate within-litter variances, yielding as many variances as there are litters per sample, and to compare the within-sample-distributions of these variances using a t-test or something related? How would you deal with this? Thank you and best wishes Louis Louis Boell MPI für Evolutionsbiologie August-Thienemannstr.2 24306 Plön [email protected] [email protected] ------------------------------------------------------------------------ PC Entertainment, wann und wo Sie wollen. Jetzt erleben. <http://redirect.gimas.net/?n=M0902FTPEntertainment> -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
