-------- Original Message -------- Subject: CVA and PCA Date: Thu, 4 Sep 2008 03:27:22 -0700 (PDT) From: Guillaume COLOMBEAU <[EMAIL PROTECTED]> To: [email protected] Dear Morphometricians I have a problem with PCA and CVA interpretations. My study deals with heterogeneous samples of Canis specimens described with cranial 3D landmarks. I did a PCA on procruste coordinates, then a CVA on coordinates in the tangent space. As example, this is a similar case to the study of P. Gunz & K. Harvati, 2007, JHE (Neanderthal 'chignon') : the number of modern individuals represents 60 to 70% of the total corpus. I understand the words of James Rohlf (in the 'Two group CVA plot' topic) that explains that CVA = PCA in standardized space, and that CVA computed on all PCs is a rigid rotation (in 'CVA and MANOVA'). That's why now I doubt: I did a CVA (with PCs summing 95% of the total variance) to evaluate the strengh of within-group variances on the calculation of the PCs. As in Gunz & Harvati 2007, scatterplots from PCA and CVA are similar, showing a rotation and a better segregation on CVA plot. How can one 'read' the strengh of within-group variances and then correctly read the PCA results? Thank you for your help Best regards guillaume -- Guillaume Colombeau PACEA-IPGQ/UMR 5199 du CNRS Avenue des Facultés Bâtiment de géologie B 18 33405 - Talence Cedex -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
