----- Forwarded message from henry wallace
<[email protected]> -----
Date: Tue, 30 Jul 2013
16:56:36 -0400
From: henry wallace
<[email protected]>
Reply-To: henry wallace
<[email protected]>
Subject: MorphoJ and procrustes
ANOVA
To: [email protected]
Hello
to all,
My questions is about using the MorphoJ to implement the
procrustes ANOVA.
I have need to find measurement errors in my
data. This is step 1. In the end, I need to discern a difference, or no
difference, between two groups. This is step 2.
For step 1--finding
measurement errors.
I have for each specimen 2 photographs
digital. Each photograph was landmarked two times. This following the guidelines
of MorphoJ User Manual page. Specimens are flat (fish) and is a 2 dimensional
analyses.
With object symmetry selected, "Individuals" are the
specimens, and "Error 1" is for image variation. "Error 2" is for landmarking
variation. Specimens are fixed. Image and landmarkings are random. Any residual
left over after effects are estimated is due to within group variations beyond
what myself has introduced into the data. Is this correct in
interpretation?
For step 2-- finding
differences.
According to Users Manual page, I can select a
second fixed main effect. This can be my grouping of specimens, no? This will
tell me if groups are significantly differentiated above expectation given
within group variation, no?
If yes, is the comparison "corrected"
for random variations in image and landmarkings? I ask because Users Manual says
that output contains random factor matrices, one of which has "The variance and
covariance components...corrected for 'lower-level' effect." Is this correct in
interpretation?
To clarify previous question, if I export this
outputted individual matrix to another softwares, such as SPSS, may I use it to
perform ANOVA to t-test for my two groups, without needing to worry about
measurement error effects?
If the answer is no, I cannot interpret
the individual matrix output from procrustes ANOVA this way, may I simply
average my data within individuals and then perform t-test in SPSS or the
like?
To clarify previous question, may I use MorphoJ Procrustes
ANOVA to only measure the measurement errors in my own data, which I
subsequently subject to averaging within specimen and finally a t-test for 2
groups in another software?
--Henry
Beloit
College
----- End forwarded message
-----
