----- Forwarded message from Øyvind Hammer <oyvind.ham...@nhm.uio.no> -----
Date: Sun, 25 Aug 2013 03:58:29 -0400
From: Øyvind Hammer <oyvind.ham...@nhm.uio.no>
Reply-To: Øyvind Hammer <oyvind.ham...@nhm.uio.no>
Subject: Re: Number of relative warp scores
To: morphmet@morphometrics.org
In this case, the number of principal components (RWs) is limited by
the number of specimens. Even if your 10 points are embedded in a very
high-dimensional space, there will be a 9-dimensional subspace that will
completely contain them. If you had 2 points, they would lie on a
straight line (1D). If you had 3 points, they would lie on a plane (2D).
Etc.
It is the job of the principal components analysis to identify the
orthogonal vectors (principal components) that span this 9-dimensional
space.
Oyvind Hammer
Natural History Museum
University of Oslo
On 2013-08-25 09:35, morphmet_modera...@morphometrics.org wrote:
> ----- Forwarded message from Jonas Kuppler
> <jonas.kupp...@uni-duesseldorf.de> -----
>
> Date: Fri, 23 Aug 2013 08:41:54 -0400
> From: Jonas Kuppler <jonas.kupp...@uni-duesseldorf.de>
> Reply-To: Jonas Kuppler <jonas.kupp...@uni-duesseldorf.de>
> Subject: Number of relative warp scores
> To: morphmet@morphometrics.org
>
> Hello all,
>
> Currently I working on intraspecific varianz of flower shape and I
> am quite new to it.
> I have 10 flowers with 220 landmarks (how I got them see below)
> each. And I used tpsRelw to calculated the relative warps (RW).
> As far as I understood I should get 436 RWs (220*2 - 4), but when I
> save the "relative warp scores", and also in the report, I only get 9
> RWs. This number seems quite low. Does anyone know why I only get 9
> RWs? Or if I did a mistake?
> If you need additional information please ask.
>
> Thank you,
>
> Jonas
>
> How I got my landmarks:
> It is a flower with 4 Petals and I have 8 Landmarks and 4 curves
> with 55 Semilandmarks. One "Start" and One "End" Semilandmark is the
> same as a Landmark.
> Then I used tpsUtil to equal landmarks and semilandmarks ("append
> tps Curve to landmarks") and delete the Semilandmarks which are
> overlapping with Landmarks (tpsUtil: "Delete/reorder landmarks").
> So I have 220 landmarks. So I created a "sliders file" for tpsRelw
> And I calculated the relative warps with tpsRelw.
>
> *
> *
> *
>
> --
> Jonas Kuppler, PhD student
> Institute of Sensory Ecology
> Heinrich-Heine-University
> Duesseldorf - Germany
> E-Mail: jonas.kupp...@uni-duesseldorf.de
>
> ----- End forwarded message -----
----- End forwarded message -----
