----- Forwarded message from Abby Drake <abbygracedr...@googlemail.com> -----
Date: Sun, 16 Jun 2013 14:26:37 -0400
From: Abby Drake <abbygracedr...@googlemail.com>
Reply-To: Abby Drake <abbygracedr...@googlemail.com>
Subject: Re: how to do shape comparison
To: morphmet@morphometrics.org
Dear Eloise,
I don't have much time but here's a really quick version of how to
compare shapes in Morphoj. I hope it helps and please feel free to
contact me if anything is unclear.
Combine your datasets in morphoj - the combine command is under the
preliminaries menu. then run a procrustes fit on the combined data.
Next make a classifiers file where one column contains the group
designation.
The file should have the same structure as in the following example:
ID Group Sex
specimen1 mating m
specimen2 nonmating f
etc.
You want one column for the file names and then columns for each
grouping variable. Make this file in excel and save it as a csv file.
highlight your dataset in the project tree and then go to the file
menu and import classifiers. go to preliminaries and edit classifiers
to check them. now go to preliminaries and find outliers. in the box
on the lower left highlight the first specimen and look at the red
lines above - do any landmarks look switched or is it in someway
digitized incorrectly? look at the first ten or so specimens for
errors. then go back to the project tree, highlight the dataset and
then go to preliminaries and generate covariance matrix, run a PCA on
the covariance matrix. right click on the pc scores and color the data
points by the group classifier. this will give a you a visual
representation of the groups and whether they are different. next run
a CVA on the data with a permutation test, in the results you will get
the procrustes distance and a p-value as well as the mahalanobis
distance and a p-value.
finally you should check for allometry. run a regression of procrustes
coordinates on log centroid size. you can color the data points on the
regression by group. under the results tab it will give you a p-value
for the slope and a % variation explained by size. i would also run a
PCA on the residuals from the regression to look at nonallometric
shape variation (make a covariance matrix from the regression
residuals). Also run a CVA on the regression residuals.
Best Wishes,
Abby
On Sun, Jun 16, 2013 at 12:33 PM, <morphmet_modera...@morphometrics.org> wrote:
>
> ----- Forwarded message from Eloise Cave <ec...@my.fau.edu> -----
>
> Date: Sun, 16 Jun 2013 00:54:09 -0400
> From: Eloise Cave <ec...@my.fau.edu>
> Reply-To: ec...@my.fau.edu
> Subject: how to do shape comparison
> To: morphmet@morphometrics.org
>
> Greetings,
> I am currently using TPS and MorphoJ software to try to do a shape
> comparison on stingray teeth between mating season and non mating season.
> What would be the procedure to show results that the two seasons are similar
> or different from one another? I have photos and landmarks on TPS files.
> currently mating season and non mating season are two completely different
> files. Should I have them as one file or should I keep them seperate? should
> I load the files into morpho J separately or as one? Currently I have them
> in Two different data sets in MorphoJ but under the same project tree. How
> can I compare those 2 data sets to show if there is a difference or not? I
> have done a PCA test for each and I know I can either use CVA or DFA but how
> or which to use I am not quite sure. A somewhat step by stem procedure and a
> small explanation would be greatly appreciated if possible. I am have a hard
> time understanding the correct procedure of morphoJ since I am new to the
> software. Oh and I'm an undergrad.
>
> Thank you
> Eloise Cave
> Florida Atlantic University
> ec...@fau.edu
>
>
>
> ----- End forwarded message -----
>
>
>
--
Abby Grace Drake, PhD
Biology Department
Skidmore College
Phone: 508.981.2783
Skype: abby.drake
Website: http://www.skidmore.edu/academics/biology/adrake.htm
----- End forwarded message -----
