Oliver,
The number of PCs is less than the number of landmarks because scale,
translation and rotation are 'removed'
I hope this helps but if not please e-mail me on my private e-mail
[EMAIL PROTECTED]
For other morphometers .. a new version of morphologika - morphologika2
is available for download at www.york.ac.uk/res/fme
And I anticipate posting a further version later this week that was
modified during the Vienna EVAN workshop this july.
The new release as of later this week will add to morphologika
1. form space (size and shape)
2. multivariate regression of shape onto an independent variable
3. input of .tps files and conversion of these to morphologika files
4. output of all results to .csv files for further analysis by standard
software
Paul O'Higgins
-----Original Message-----
From: morphmet [mailto:[EMAIL PROTECTED]
Sent: 02 August 2006 15:04
To: [email protected]
Subject: output file in morphometrika
My e-mail for replies: [EMAIL PROTECTED]
Dear list members:
using the software morphologika I came across a problem regarding the
evaluation of the variables, which determine the different PCs in PCA.
In total, I set 26 3D landmarks, but in the output file there are only
75 (instead of 78) variables listed. Hence, I was wondering, what these
different variables actually mean. Am I right that variable 1 is the x-
coordinate of landmark 1, variable 2 = y-coordinate of lm 1, variable 3
= z-coordinate of lm 1, variable 4 = x-coordinate of lm 2, variable 5 =
y-coordinate of lm 2, and so on?
I actually would like to assess the influence of each landmark on each
PC by inspecting the eigenvectors of each variable for that component.
However, since I am not sure what the listed 75 variables actually mean,
I would like to ask any of you to give me an advice. Furthermore, I was
wondering whether it is possible to use the magnitude of the listed
correlation coefficient between both the variable and the PC to see,
which variables explain most of the respective PCs. In conventional PCA
it is recommended (as a rule of thumb) to consider all the correlations
> 0.5 as important and to discard all the variables in the
inerpretation, which have lower correlations with the respective PC. I
was wondering, whether this criterion is also applicable here.
I will be very glad if anybody can give me an advice on how to proceed.
I'll summarize the replies and post them to the mail server.
My e-mail for replies is: [EMAIL PROTECTED]
My best wishes and thanks,
Oliver Betz
Zoology Dptmt.
University of Tübingen
Germany
--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org
--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org
