A quick cautionary note on using Procrustes shape
variables in statistical software which is not
specific to this type of data: if shape
coordinates are used, parametric tests may
produce wrong results when the software does not
'see' that shape coordinates are redundant
because of the loss of df in the superimposition
plus, if one has sliding semilandmarks, those lost in the sliding process.
This is why I tend to use PCs: all of them if
possible; a subset of the first ones, if I can't
include them all. If a subset of the first ones
is used, one has to be careful in convincingly
showing that they preserve the greatest majority
of information in the original data and maybe has
to do some sensitivity analysis to the inclusion
of more or less PCs. I would suggest some
sensitivity analysis also to the
inclusion/exclusion of smallest samples, if
sample size is very heterogeneous across samples (e.g. in DA/CVA).
With PCs of shape data what I would definitely
avoid is stepwise procedures and analyses of PCs
one at a time. Adams et al., 2011, Journal of
Human Evolution exemplifies why this is not a
good idea and has refs to previous papers making the same point.
If one can avoid dimensionality reduction by
using, for instance, resampling methods based on
distances, that might be a better option than
reducing dimensionality. I fear that these
methods are not yet available for all types of
analyses (e.g., in my outdated experience, many
of those used in spatial data analyses). However,
there's an increasing abundance of distance-based
resampling methods especially in R (quite a few
recently added to the list by Dean Adams et al.).
Besides avoiding potential loss of information,
when these methods are using Procrustes
distances, they preserve the geometry of the
shape space produced by the superimposition.
Good luck.
Cheers
Andrea
At 22:59 10/02/2015, Emma Sherratt wrote:
Gabi,
While Michael is correct i how you can export
the CV scores from MorphoJ, I would highly
recommend against exporting the CV scores to
plot against other parameters. The reason being
that CVA should not be used like Principal
Components Analysis. CVA axes should be used for
inspecting the data for the aspects of shape
that delimit and discriminate between two or
more groups. Not as reduced axes for use in correlation tests.Â
Emma
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Emma Sherratt, PhD.
Lecturer in Zoology,
Zoology Division, School of Environmental and Rural Science,Â
Room L120 Bldg C02,Â
University of New England,Â
Armidale, NSW, Australia, 2351
Tel: +61 2 6773 5041
email:Â <mailto:[email protected]>[email protected]
Caecilians are legless amphibians...
                     __
   (\  .-.  .-.  /_")
    \\_//^\\_//^\\_//
  `"`  `"`  `"`
learn more about them here:
<http://www.emmasherratt.com/caecilians>www.emmasherratt.com/caecilians
On 11 February 2015 at 08:33, gnavas
<<mailto:[email protected]>[email protected]> wrote:
Dear Morphometrics Wizards,
I have 2 questions for which I am hoping to get help on:
Question 1. CVA analysis: I have 6 sites. Within
each site I have at least 22 samples. When I ran
a CVA comparison on these 6 sites in Morpho J. I
was hoping to find a way to get an actual value
for each of my samples that plotted. So, for
instance, CV1 values for each specimen, and the
same for CV2. In the results tab, I can only
find data relevant to my landmarks and canonical
variates coefficients relating to lanmarks.
I would love to figure out a way to export my
CV1 and CV2 values for each specimen to then
plot that against other parameters that may be
effecting shape at those 2 or more canonical
variate axes. Has anyone run into this problem, and found a solution?
Question 2. Discriminant Analysis:Â
Again, 6 sites with 22 samples, but this time I
was only able to compare 2 sites at a time. Has
anyone ever been able to run 1 site against all
remaining sites? It would be great to get an
idea of how my sites compare to all others,
rather than to just one other at a time. I
suspect I have to play with my classifier
variables, but I am not sure how to go about
that. At this point, I have made 1 classifier
variable that allows me to distinguish the different sites.
If any of you have run into this or simply know
how to do this, please let me know. I am also
happy to give more detail on my study if that would help?
Thank you in advance!
Gabi
--
MORPHMET may be accessed via its webpage at
<http://www.morphometrics.org>http://www.morphometrics.org
To unsubscribe from this group and stop
receiving emails from it, send an email to
<mailto:[email protected]>[email protected].
--
MORPHMET may be accessed via its webpage at
<http://www.morphometrics.org>http://www.morphometrics.org
To unsubscribe from this group and stop
receiving emails from it, send an email to
<mailto:[email protected]>[email protected].
Dr. Andrea Cardini
Researcher in Animal Biology, Dipartimento di
Scienze Chimiche e Geologiche, Università di
Modena e Reggio Emilia, l.go S. Eufemia 19, 41121 Modena, Italy
Adjunct Associate Professor, Centre for Forensic
Science , The University of Western Australia, 35
Stirling Highway, Crawley WA 6009, Australia
E-mail address: [email protected], [email protected]
WEBPAGE: http://sites.google.com/site/hymsfme/drandreacardini
Summary of research interests at:
http://www.dscg.unimore.it/site/home/ricerca/aree-di-ricerca/evolution-taxonomy-and-forensics.html
FREE Yellow BOOK on Geometric Morphometrics:
http://www.italian-journal-of-mammalogy.it/issue/view/405
or full volume at:
http://www.italian-journal-of-mammalogy.it/public/journals/3/issue_241_complete_100.pdf
Editorial board for:
Zoomorphology:
http://www.springer.com/life+sciences/animal+sciences/journal/435
Journal of Zoological Systematics and
Evolutionary Research: http://www.wiley.com/bw/journal.asp?ref=0947-5745&site=1
Hystrix, the Italian Journal of
Mammalogy: http://www.italian-journal-of-mammalogy.it/
--
MORPHMET may be accessed via its webpage at http://www.morphometrics.org
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
