Dear Callie,
I am not sure I fully understand your question. Perhaps you could
provide details as to how your 20 projectile points are transformed into
data, how the pairwise rotation of two configurations was applied to
your 20 projectile points, and what is the aim of the test.
Now, let's assume for the sake of argument that you have converted your
projectile points into data as a set of n 2D landmarks (in the case of
circle you would have used the same amount of points to form a circle).
Let's also assume that you have used as starting point for your PROTEST
two of these configuration of landmarks (i.e., the matrix with n rows
and 2 columns containing the x,y coordinates of each landmark; one
matrix for one projectile point, the other for another projectile point).
If the above were true and based on the way the procedure works (in my
case, I'm referring to the description in Peres-Neto & Jackson 2001 -
Oecologia), I think a reasonable expectation, for two quite similar
shapes, would be:
- that after Procrustes superimposition, the x,y coordinates of
corresponding points/landmarks would be extremely similar, giving a very
low "Procrustes sum of squares (m12)"
- that randomly exchanging the order of the points in one of the two
configurations would basically never give you a "Procrustes sum of
squares (m12)" as small (because, in lay terms, if in one of the two
landmark configurations you exchange the position of two or more of the
landmarks, the two shapes will not match as precisely as your original
configuration)
Blake has given you a potential explanation of why a specific value,
perhaps my hypothetical scenario (if it applies) can give you a hint on
why that is always the same.
If the hypothetical scenario above were correct, I would see no special
issue with the value being always the same (and extremely low). In that
case, however, there might be an issue with the practical utility of
performing the test in the first place.
I hope this helps,
Carmelo
Il 16/01/2018 7:51 PM, Callie Diduck ha scritto:
Hi,
I am wondering if any one can help me. I have been running a Procrustes
Rotation of Two Configurations and PROTEST using the vegan package in R.
The Procrustes seem to be working but I am having an issue with the
results of the PROTEST and the correlation values.
I am getting identical p values every time I run the script. I have a
sample size of 20 projectile points and every time I run the statistics
my p value is
equal to 0.00009999. I ran a test and I compared a circle to one of my
projectile points and I still got a p value of 0.00009999.
Here are the results for the test of the projectile point compared to a
circle:
Procrustes sum of squares:
1.817e+04
Call:
protest(X = p730, Y = pTest, permutations = 10000)
Procrustes Sum of Squares (m12 squared): 0.06296
Correlation in a symmetric Procrustes rotation: 0.968
Significance: 9.999e-05
Permutation: free
Number of permutations: 10000
If anyone has any idea of why this could be happening or how to fix it
that would be greatly appreciated.
Thanks
Callie Diduck
--
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.
