-------- Original Message --------
Subject: Re: [Fwd: Re: Morphometrics of small, variable specimens--embryos]
Date: Wed, 30 Mar 2011 18:09:30 -0400
From: [email protected]
To: [email protected]
Andrea,
Thanks for this.
Regarding my second post, I can restate the issue more succinctly.
I correct the data by regressing the Procrustes coordinates on centroid
size. I further correct the data for developmental stage by regressing the
centroid size residuals on somite number.
This should remove variation correlated with size and with developmental
stage, at least to the degree a linear model will allow.
The PC1 scatter of the data are such that early and late stage embryos
cluster together toward one end and intermediately staged embryos are
scattered toward the other.
How interpretable is the wireframe deformation in a case like this, where
at one extreme of the major linear trend both early and late staged
embryos are found together. The shape change along PC1 relates to
ontogeny, but not in a simple straight forward way. The earlier and late
staged embryos are very different from on another, and this can be seen
just by eye, yet they occupy the same region of PC1. PC2 is similar in
this pattern but not as severe.
Do these observations say that there is still a strong non-linear
component of variation in my data that the above corrections leave behind?
Is it correct to interpret a wireframe scaled to the area occupied by both
developmentally old and young embryos as a composite of features typical
of each stage?
Eric
>
-------- Original Message --------
Subject: Re: [Fwd: Re: Morphometrics of small, variable
specimens--embryos]
Date: Wed, 30 Mar 2011 03:52:13 -0400
From: andrea cardini <[email protected]>
To: [email protected]
A quick comment on errors, Eric, besides what David told you.
By repositioning the specimen you will be able to get also that source of
error in the model. Then, you can test the total error or do the test
after
partitioning it between positioning and digitizing error, as Daid
suggested
and can be easily done in MorphoJ. However, this won't do much to tell
you
how good the 2D approximation of your 3D structures is. To add that you
would need 3D measurements (at least on a subsample of specimens). You
can
possibly try to minimize this error by choosing almost coplanar
landmarks.
About your second question, I read the message quickly and I am not sure
about what exactely you're doing. Generally, if your aim is to compare
groups by holding the effect of a covariate constant, a fairly
traditional
approach would be a MANCOVA (or its version using resampling stats).
There's an example about this in the context of allometry in the help
file
of TPSRegr (test for common slope and homogeneity of intercept) and the
method is related to MorphoJ's 'size-correction', I believe.
Good luck.
Cheers
Andrea
At 12:13 29/03/2011 -0400, you wrote:
-------- Original Message --------
Subject: [Fwd: Re: Morphometrics of small, variable specimens--embryos]
Date: Tue, 29 Mar 2011 12:06:03 -0400
From: [email protected]
To: [email protected]
---------------------------- Original Message
----------------------------
Subject: Re: Morphometrics of small, variable specimens--embryos
From: "P. David Polly" <[email protected]>
Date: Tue, March 29, 2011 10:02 am
To: [email protected]
--------------------------------------------------------------------------
You might get part of the error by removing one of the PCs, but that
approach is less precise than partitioning it out. The PC axes are
simply
axes of greatest variation so they are not directly associated with any
causal process (including mounting error). Error due to mounting may be
concentrated on one PC axis, but it may be spread across more than one,
plus
non-error may also contribute to the same PC axis (e.g., true bilatral
asymmetry might logically contribute to shape variation in exactly the
same
way as error in the saggital section).
The ANOVA approach is equivalent to regressing out the error (ANOVA and
regression are in some senses the same thing, one for factor variables
and
the other for continuous variables). With the ANOVA you know you are
removing only variation due to the mounting, and you also know you are
removing most of it (to the extent that two mountings per specimen are
representative of the amount of the total amount of error).
----- Original Message -----
From: <[email protected]>
To: "P. David Polly" <[email protected]>
Sent: Tuesday, March 29, 2011 11:55 AM
Subject: Re: Morphometrics of small, variable specimens--embryos
Thanks! What about regressing the data on a PC that appears to
explain
mounting errors?
Eric
Hi Eric,
I think you're on the right track with mounting the same specimen
more
than
once. If you do every specimen two or three times you can partition
out
the
shape variation due to mis-alignment of the plane. To do this you add
an
additional level to your ANOVA so that it has factors for between
group,
between individual, and between mount variation. There are several
papers
you can cite for this method, but my favourite is Baily and Byrnes,
1990.
A
new, old method for assessing measurement error in both univariate
and
multivariate morphometric studies. Syst. Zool. 39: 124-130.
Best wishes,
David
-----------------------
Dr. P. David Polly
Department of Geological Sciences
Indiana University
1001 E. 10th Street
Bloomington, IN 47405 USA
[email protected]
+1 812 855 7994
http://mypage.iu.edu/~pdpolly/
(Adjunct in Biology and Anthropology)
----- Original Message -----
From: "morphmet" <[email protected]>
To: "morphmet" <[email protected]>
Sent: Tuesday, March 29, 2011 11:36 AM
Subject: Morphometrics of small, variable specimens--embryos
-------- Original Message --------
Subject: Morphometrics of small, variable specimens--embryos
Date: Tue, 29 Mar 2011 10:28:14 -0400
From: [email protected]
To: [email protected]
Hello all,
I am currently doing 2D and 3D analyses of midgestational mouse
embryos.
My sample is variable owing to ontogenetic variation. Genetic
variation
is
very minimal, as my strains are mostly congenics, having practically
identical genetic backgrounds but differing only at one or two loci.
The
specimens are also very small.
First, for the 2D analysis, I am photographing freshly harvested and
unfixed embryos in three different orientations (top, lateral, and
"frontal" = palatal view), mounting in a petri dish of cold saline
and
photo'ing two separate times per view. Each set of images per
specimen
is
landmarked twice, and all the data will be subject to an initial
procrustes ANOVA to assess the relative strengths of the different
effects
of mounting, landmarking, genotype, and specimen. However, I expect,
and
experience shows, that a significant portion of the variance in the
data
will be due to mounting errors. The embryos are small and difficult
to
position. GPA will take care of rotational errors. But slight
rotations
out of the plane (pitch and yaw) will produce variation in the data
that
will look like shape variation. My hope is that by mounting and
photo'ing
twice, I will reduce pitch/yaw errors.
Will the mean square of the mounting effect reflect the amount of
those
types of errors?
If I can identify a PC that appears to capture pitch or yaw, can I
regress
the procrustes coordinates on that PC in order to remove those
errors
from
the data?
Thanks.
Eric
[email protected]
University of Calgary
Faculty of Medicine
Dr. Andrea Cardini
Researcher in Animal Biology
Dipartimento di Biologia, Universitá di Modena e Reggio Emilia, via Campi
213, 41100, Modena, Italy
tel: 0039 059 2055017 ; fax: 0039 059 2055548
Honorary Fellow
Functional Morphology and Evolution Unit, Hull York Medical School
University of Hull, Cottingham Road, Hull, HU6 7RX, UK
University of York, Heslington, York YO10 5DD, UK
Adjunct Associate Professor
Centre for Forensic Science , The University of Western Australia
35 Stirling Highway, Crawley WA 6009, Australia
E-mail address: [email protected], [email protected],
[email protected]
Webpage: http://sites.google.com/site/hymsfme/drandreacardini
Datasets:
http://ads.ahds.ac.uk/catalogue/archive/cerco_lt_2007/overview.cfm#metadata
