Hi,
it's hard to give clear answers as context is missing. But I will try to
give my general opinion on this.
If you had some a priori idea of meaningful shape variation and you were
sure this variation is represented by an individual PC axis (and that
axis only), it would be sensible to make that computation for that axis.
However, in your case - and in most cases where a similar question is
posed - this doesn't seem to apply (i.e., from the way you ask the
question, it sounds like you want to compute variance for each of the
axes separately simply because PCA gives you a set of orthogonal axes).
Then - provided that I don't know enough about your
system/question/analysis and this is just a general opinion - I would
just stick to multivariate measures of variation.
I hope this helps.
Carmelo
--
==================
Carmelo Fruciano
Institute of Biology
Ecole Normale Superieure - Paris
CNRS
http://www.fruciano.it/research/
On 16/11/2018 19:45, colymboides wrote:
Hi everyone,
I extracted Procrustes shape variables describing 95% of the variation
(PC1-PC4) in my data and plotted shape changes in morphospace. By
plotting the warps I know what shape changes are described along each PC
axis. For example, PC1 describes elongated (PC1 min) to stout (PC1 max)
shapes. Would it be sensible to calculate some measure of variance/morph
disparity among my species using the PC scores for each PC axis (e.g.
compare species X PC1 to species Y PC1) as opposed to calculating
disparity by using combined PC axes?
Nick
--
MORPHMET may be accessed via its webpage at http://www.morphometrics.org
---
You received this message because you are subscribed to the Google
Groups "MORPHMET" group.
To unsubscribe from this group and stop receiving emails from it, send
an email to morphmet+unsubscr...@morphometrics.org
<mailto:morphmet+unsubscr...@morphometrics.org>.
--
MORPHMET may be accessed via its webpage at http://www.morphometrics.org
---
You received this message because you are subscribed to the Google Groups "MORPHMET" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to morphmet+unsubscr...@morphometrics.org.