-------- Original Message --------
Subject: Query about CVA in MorphoJ
Date: Thu, 7 Apr 2011 21:43:50 -0400
From: ITO Tsuyoshi <[email protected]>
To: [email protected]
Dear all,
Canonical variate analysis (CVA) is a method used to find the shape
features that best distinguish among multiple groups of specimens.
To my knowledge, CVA requires that the number of specimens in each group
should be larger than the number of independent variables.
MorphoJ may perform CVA even if the number of landmarks exceed the
number of specimens of each group.
My question is what is done in such a case (Are independent variables
automatically reduced by using PC scores)?
Is it plausible to perform CVA in which the number of landmarks is 10
(hypothetically speaking) while the number of specimen for a group is one?
In relation to this, does any software can assignment test in which the
CVA axes determined by a number of known groups to assign an unknown
specimen to one of the known groups (using Mahalanobis distance) based
on 3D landmarks?
Thank you very much for your help.
Best regards,
Ito
ITO Tsuyoshi
Primate Research Institute, Kyoto University
[email protected]
