-------- Original Message --------
Subject:        How to test landmarking accuracy
Date:   Thu, 20 Oct 2011 11:10:30 -0400
From:   Wechsler, Samuel <[email protected]>
To:     [email protected] <[email protected]>



Dear all

I am completely new to the morphometrics business. For my master thesis
I work with pictures of brown trout. I would like to check how accurate
I work when a) I set the landmarks and b) when I take pictures. To check
this I landmarked three different sets of pictures:

1^st set: 5 times exactly the same picture (to test how accurately I can
landmark)

2^nd set: 5 times exactly the same fish on 5 pictures taken on different
days (to test how accurately I take photographs)

3^rd set: 5 different fish

I am looking for a statistical measure that can inform me about the
spread of my landmarks around their respective average (something
equivalent to standard deviation or alike). When I do the procrustes fit
in MorphoJ the results section (attached below) provides “Procrustes
sums of squares”.

Now finally here my questions:

1)What are these Procrustes sums of squares exactly? Are they an overall
sum of all squared procrustes distances over every landmark and every
observation (1 observation = coordinates from 1 picture)?

2)Can the Procrustes sums of squares be converted/standardized to a
measure that can be compared between sets of pictures with differing
number of observations?

Thank you for your help.

Kind regards

Sämi

PS: Below an example result output from morphoJ

Dataset 'NM007 (5x)'

16 landmarks in 2 dimensions.

The dataset contains 5 observations, of which 5 are included for analyses.

Average shape:

Lmk.             Axis 1 (x)    Axis 2 (y)

   1                  -0.29489099                   0.01139803

   2                  -0.26145670                   0.02011783

   3                  -0.24846114                   0.02003904

   4                  -0.23435917                   0.02014906

   5                  -0.22792309                 -0.03741346

   6                  -0.12578189                 -0.00296614

   7                  -0.12529903                   0.07369970

   8                  -0.12574654                 -0.07027571

   9                  -0.14574592                 -0.04920278

10                   0.04206763                   0.08215905

11                   0.28577462                   0.05461450

12                   0.35694734                   0.04214234

13                   0.40457048                   0.00679023

14                   0.35382919                 -0.02944051

15                   0.24278591                 -0.06530739

16                   0.10368929                 -0.07650380

Procrustes sums of squares: 0.0016642169841469471

Tangent sums of squares: 0.0016636098057431449

Data matrices in this dataset:

    - NM007 (5x), raw data

    - NM007 (5x), centroid size

    - NM007 (5x), Procrustes coordinates

Reply via email to