-------- Original Message --------
Subject: Re: multiple comparisons test for Procrustes ANOVA
Date: Fri, 12 Sep 2008 01:59:12 -0700 (PDT)
From: Chris Klingenberg <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Organization: University of Manchester
To: [email protected]
References: <[EMAIL PROTECTED]>

Dear Shane

The simplest way to do the comparison of the amounts of FA in different
groups, to my knowledge, is to use an analogue to Levene's test, which
has been recommended for univariate distance data in traditional FA
analyses (e.g. Palmer, A. R., and C. Strobeck. 2003. Fluctuating
asymmetry analyses revisited. Pp. 279–319 in M. Polak, ed. Developmental
instability: causes and consequences. Oxford University Press, New York.)

For geometric morphometrics, Leandro Monteiro and I have suggested such
a method. For each individual, you compute a scalar measure of asymmetry
(corrected for directional asymmetry), which you can then compare with a
t-test or univariate ANOVA according to your experimental design.

As measures of asymmetry, there are two choices: Procrustes distance and
a measure based on Mahalanobis distance. These are measures of absolute
shape differences and shape differences relative to the variation in the
total sample, respectively. In practice, both distances tend to give
similar results (in my experience -- but this presumably depends on
study organisms and experimental settings). More detail is provided in
the following paper:

Klingenberg, C. P., and L. R. Monteiro. 2005. Distances and directions
in multidimensional shape spaces: implications for morphometric
applications. Systematic Biology 54:678–688.
http://www.flywings.org.uk/PDF%20files/SystBiol2005.pdf

This method is implemented in MorphoJ. If you run a Procrustes ANOVA, an
output dataset will be generated, which contains the two measures of
asymmetry for every individual in your sample. You can export these
values to a text file (File > Export dataset) and use a univariate ANOVA
to test for differences among groups, according to the design of your study.
For details:
http://www.flywings.org.uk/MorphoJ_guide/frameset.htm?variation/procANOVA.htm


A fairly large-scale application of this approach was used in a study
that related the amouts of variation and FA in size and shape across
115 groups of flies with distinct genotypes:

Breuker, C. J., J. S. Patterson, and C. P. Klingenberg. 2006. A single
basis for developmental buffering of Drosophila wing shape. PLoS ONE
1(1): e7.
http://www.flywings.org.uk/PDF%20files/PLoSONE2006.pdf

I hope this helps.

Best wishes,
Chris


morphmet wrote:


-------- Original Message --------
Subject:     multiple comparisons test for Procrustes ANOVA
Date:     Wed, 10 Sep 2008 06:38:07 -0700 (PDT)
From:     Shane Welch <[EMAIL PROTECTED]>
To:     [email protected]



Hello everyone,



I'm new to geometric morphometric analysis but I have a good
understanding of univariate statistics.  I've been recruited by my lab
to perform a quick analysis for wing landmark data.  The objective is to
compare fluctuating asymmetry (FA) among different treatment groups. We
have 8 landmarks per wing (right vs. left), landmarks were recorded by
the same person 3 times, there are approximate 100 replicates per
treatment, and 4 treatment groups.  I preformed a *Procrustes fit* and a
Procrustes ANOVA.  The main effect of treatment was significant, but
here I get lost, I would like to know how to perform the equivalent of
post-hoc multiple comparisons test to determine which treatments
exhibited greater FA.  My understanding of multivariate statistics is
limited, so if the answer is in that direction, please go slow.

Any help is great.



Thanks



Shane




--
***************************************************************
Christian Peter Klingenberg
Faculty of Life Sciences
The University of Manchester
Michael Smith Building
Oxford Road
Manchester M13 9PT
United Kingdom

Telephone: +44 161 275 3899
Fax: +44 161 275 5082
E-mail: [EMAIL PROTECTED]
Web: http://www.flywings.org.uk
***************************************************************



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

Reply via email to