-------- Original Message --------
Subject: RE: CVA and MANOVA
Date: Thu, 28 Feb 2008 09:37:07 -0800 (PST)
From: [EMAIL PROTECTED]
To: [email protected]

Hi,

The complication with CVA in estimation of shape configurations is that
CVA distorts the original tangent space of shapes were parametric
analyses are often applied. There is an interesting paper discussing
this aspect:

Klingenberg, C. P. and L. R. Monteiro. 2005. Distances and directions
in multidimensional shapes spaces: implications for morphometric
applications. Systematic Biology 54(4):678-688.

However, some friends in the forum have already suggested interesting
options on how to estimate shape configurations in the context of CVA.

As for the possible problem of using multivariate vectors (relative
warps) rather than the original variables (part warps), in statistical
contrast tests (CVA-MANOVA), I think Dr. Rohlf has already made clear
how relative warps are nothing more than statistically orthogonal
(independent) linear combinations of the original variables. Hence,
there is no "non-independence data problem" if using RW for a MANOVA.
Problems may be found when you have collinearity or sphericity among
the original variables; then you will have a sort of non-independence,
something a factor analysis could resolve.


Good luck

Pablo

Pablo Jarrin
Grad student
Dept. of Biology
Boston University



--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org

Reply via email to