-------- Original Message --------
Subject: RE: Time series analysis on landmarks
Date: Mon, 19 Mar 2012 09:23:58 -0400
From: J.C. van Rijssel <[email protected]>
To: <[email protected]>
Dear Paolo,
Thank you for your reply. I will look in to that.
Best regards,
Jacco van Rijssel
-----Original Message-----
From: morphmet [mailto:[email protected]]
Sent: zaterdag 17 maart 2012 2:13
To: morphmet
Subject: Re: Time series analysis on landmarks
-------- Original Message --------
Subject: Re: Time series analysis on landmarks
Date: Fri, 16 Mar 2012 17:43:38 -0400
From: [email protected]
To: [email protected]
You should not use CVA to test differences in shape between any sort of
classes (decades, sex, species, etc.). Use, instead, Procrustes
Coordinates. You can find detailed discussion on morphmet about this topic.
Moreover, there is in the data a temporal non-independence that should
be removed when performing MANOVA or regression analysis.
Time is a high anisotropic source of covariation between observations.
Variation partitioning could deal with this situation. Or gls.
I hope this helps
Best
Paolo
-------- Original Message --------
Subject: Time series analysis on landmarks
Date: Fri, 16 Mar 2012 12:11:28 -0400
From: J.C. van Rijssel
<[email protected]>
To: <[email protected]>
Dear Colleagues,
Currently, I am working on a large dataset of fishes where I use
preserved specimens up to 30 years ago. This results in a nice dataset
where I can check if the body shapes of these fishes have changed over
the past 30 years (at three year time intervals).
After analysis (CVA),
body shape does change within a decade (concurrent with ecological
changes in the fishes' environment).
However, in the most recent years, the body shape seems to return back
to the old shape of my first year (also concurrent with the ecological
changes).
What I would like to know if there is a way to
(statistically) check
whether the body shape has really returned to its original form or that
this might be just a coincidence that my first and last year happen to
have the same shape (no significant differences in the permutation test
of the CVA)?
I think there should be some sort of time series analysis which could
solve the issue. Does anyone know of a statistical program which can do
a time series analysis (or autocorrelation) on landmarks?
Best regards,
Jacco van Rijssel
PhD-Student Integrative Zoology
Leiden University
The Netherlands
--
Paolo Piras
Center for Evolutionary Ecology
and
Dipartimento di Scienze Geologiche, Università Roma Tre Largo San
Leonardo Murialdo, 1, 00146 Roma
Tel: +390657338000
email: [email protected]
