Hai-Jun SU <[email protected]> ha scritto:
Dear all,
Glad to be a part in this group.
Dear SU,
I'm not sure I fully understand your situation so please forgive me if
I don't address your questions properly.
I am SU from Insubria Univeristy of Italy now, at present, I am trying to
use GMM to analyze the squirrel mandibular morph to distinguish three
geographic populations (to test how similar they are), at two levels, i.e.
Ind and FA. When I want to investigate if they were similar allometry
pattern using MANCOVA, I got a problem, the results of MANCOVA （both
results by SPSS and TpsRegr, and the shape data were superimposed as a
subset separately）showed extremely significant, but from the regression
plots and their slopes values( were 0.27, 0.31 and 0.34 respectively, from
fitting of regression scores and Log CS ), it seemed similar slopes and
easy to centered, so, my puzzles are:
1) Should I believe in these results? did it really mean they can't be
forced to parallel slopes?
2) If yes, does it mean that I can't make the regression for 3 populations
together in MorphoJ ( including with and without pooling within group) ,
i.e. I have to make the allometry-related regression analysis in separate
populations (superimposed separately) in the next steps?
I'm not sure which tests you exactly did. I guess that performing a
MANCOVA (testing for both the population and size effects and their
interaction) in a general-purpose statistical software (SPSS) and/or
performing the procedure suggested in the TPSRegr help (test for
common slopes) would be the way to go to know if you can perform a
single regression or not.
3) When I was using TpsRegr, I imported the NTS file got from TpsDig, but
it contained all observations of pictures, not individuals, so how can I
use TpsRegr with averaged individuals (procrustes coordinates and Log CS
values) ? what's the differences between them?
I guess the easiest, if you're using MorphoJ and you have some
identifier that identifies each specimen, is to average them in
MorphoJ, then export the averaged shapes and CS (or logCS) and finally
convert the file in a text editor to NTS format (which, given how
simple is the NTS format for use in the TPS series, is easily done).
Of course, you can also compute average shapes after superimposition
in other software (such as Excel) but this might be more time consuming.
I hope this helps.
Best,
Carmelo
--
Carmelo Fruciano
Marie Curie Fellow - University of Konstanz - Konstanz, Germany
Honorary Fellow - University of Catania - Catania, Italy
e-mail [email protected]
http://www.fruciano.it/research/
--
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].
