-------- Original Message --------
Subject: Re: ANOVA in MorphoJ
Date: Wed, 5 Oct 2011 10:26:19 -0400
From: [email protected]
To: [email protected]
Hi,
The Procrustes ANOVA function produces covariance matrices for the
individual effect as well as specified errors (such as mounting and
landmarking errors), and these matrices can be used to compare differences
in how the variation among those levels is structured. But, after a PCA
MorphoJ dosen't project the data within the subspace associated with each
ANOVA level onto to a set of vectors to make a scatter plot, like it does
in PCA's outside the Procrustes ANOVA option. Why is this?
The motivation behind the question is to use Procrustes ANOVA to remove
the error 1 and residual components from the original data, and then to
analyze those processed data; as opposed to more simply averaging the
original data within individuals and analyzing that.
Thanks. Eric.
-------- Original Message --------
Subject: ANOVA in MorphoJ
Date: Tue, 27 Sep 2011 17:21:06 -0400
From: [email protected]
To: [email protected]
Hi,
A dataset containing specimens each mounted and photo'd twice with each
image digitized twice is cranked through a procrustes ANOVA. MorphoJ
outputs matrices associated with specimen and each error. There is also a
dataset created called "individual values." If I understand the MorphoJ
website correctly, this dataset contains the average procrustes
coordinate
values for each specimen, but is this averaged dataset calculated after
the variance-covariance associated with mounting and replicate
landmarking
is removed? The PCs generated from the "individual values" output are
slightly different compared to PCs generated from the first covariance
matrix the ANOVA creates.
Thanks
Eric
