----- Forwarded message from andrea cardini <[email protected]> -----
Date: Thu, 28 Mar 2013 04:12:11 -0400
From: andrea cardini <[email protected]>
Reply-To: andrea cardini <[email protected]>
Subject: Re: combining two data set
To: [email protected]
Hi Pepijn,
thanks for your question.
I read Dean's paper long time ago. I seem to
remember he suggested at least 3 possible approaches.
One (or two) of them worked by 're-articulating'
separate structures and may require information
on the relative positions of those structures
and/or some landmarks in common between them.
For instance, when I was working on monkeys with
3D landmarks, I 're-articulated' the cranium and
mandible, which I had digitized separately. We
should have described how I did it in a paper
(ref. below) but basically I took advantage of
three common points I had digitized on the
disarticulated mandible (together with all its
landmarks) and on the mandible articulated on the
cranium (together with the cranial landmarks).
Those three points (as they were not part of the
real configurations we were interested) were later discared.
Paul O'Higgins has a software for registering
different views/structures with this method, and
I think also the NYCEP has got one.
The third method suggested by Dean is much easier.
For shape, I think he simply appended the matrix
of shape variables of structure A (say the
cranium) to the matrix of shape variables of
structure B (say the mandible). The way I'd do it
would be by using PCs of shape in the two
datasets (whether all of them or after some
dimensionality reduction it depends but should
not make a big difference if dimensionality
reduction is done carefully - e.g., check if
shape distances are virtually unchanged using all
PCs or the first whatever). Using PCs has the
advantage of removing possible redundant
information (e.g., in symmetric objects etc.).
I'd tend to think that using this approach there
should not be landmarks in common, otherwise some
information is used twice in the analysis.
However, again, if it is very few common
landmarks out of many, it may not make an appreciable practical difference.
I confess I did not remember he added the ratio
of size in the two structures. I guess this was
in order to take into account somewhat their
relative sizes (as one would do in traditional
morphometrics with ratios). I suspect that this
may be an option, depending on what one precisely wants to do.
I've never published it but when I checked with
monkeys shape distances after rearticulating the
cranium and mandible or simply appending their
PCs were, I believe, virtually identical. In that
case, I did not add also the size ratio.
This approach is clearly very simple in terms of
computations. The disadvantage is that one is no
longer working in a shape space defined by a
specific landmark configuration and can't go back
to it straight away for the visualization. What I
mean is that, say, I am testing differences
between sexes in skulls (cranium + mandible). If
I find that they're significant for shape using
the 'rearticulated' cranium and mandible (the
approach I used on monkeys), I can visualize
sexual dimorphism using the male and female mean
shapes of the whole skulls. However, if I am
using the 'appended matrices approach', I will
have to show separately the mean mandible and the mean cranium.
I hope this helps.
Dean will certainly give you better advice and correct my poor explanations.
Cheers
Andrea
PS
One last thing on the 'rearticulated approach'.
In 2D (pictures) that's pretty simple with a few
landmarks in common in different structures
photographed in the same view. By this I mean,
for instance, side view of the left side of the
cranium and side view of the left hemimandible.
Of course, there will be some issues with how the
flattening of the third dimension is affecting
the information in one or the other structure and
in relation to their relative position in the
'real' fully articulated 3D skull.
However, if you have ventral, side and dorsal
view of the cranium (same structures, just
different views) in 2D, even with landmarks in
common, there might be some really serious
problems in rearticulating the different views
because what you're actually doing is trying to
reconstruct 3D data from 2D ones. That's more
difficult. Fadda and Corti (or Corti and Fadda)
did it with a triangulation method in an old
paper. I can send you the pdf, if you want,
as it is difficult to find. For a case like
this, if you're unable to do an accurate
reconstruction in 3D starting from 2D, Dean's
'appended matrices approach' may be an
interesting and simple option. Bear in mind, all
the information in the relative positions of
those 3 views will be lost, but you'll keep the 'within-view' information.
Fairly recently there was also a paper in
Palaeontologica Electronica (can't find the ref.
right now, sorry) suggesting basically a more
sophisticated version of Fadda and Corti's
approach using, I believe, free software which
allows to estimate 3D images from many 2D ones
taken following a given protocol.
Cardini A., Elton S., 2008 -
<http://docs.google.com/fileview?id=0B7MW9En7Pb3VN2E5YzI5OTYtYzRmYi00YmQ0LWE5NDItYjVkYzUyZjIxYTJi&hl=en_GB>Does
the skull carry a phylogenetic signal? Evolution
and modularity and in the guenons. Biological
Journal of the Linnean Society, 93: 813–834.
Reprint at: https://sites.google.com/site/hymsfme/drandreacardini/publications
At 05:48 28/03/2013, you wrote:
>----- Forwarded message from Pepijn Kamminga <[email protected]> -----
>
>Date: Wed, 27 Mar 2013 07:23:12 -0400
>From: Pepijn Kamminga <[email protected]>
>Reply-To: Pepijn Kamminga <[email protected]>
>Subject: Re: combining two data set
>To: [email protected]
>
>Dear Andrea,
>Â
>I have a similar problem in combining 2
>datasets. Do you mean by combining the shape
>variables in xls that you combine the landmark
>coordinates from 1 individual in the first
>dataset with the landmark coordinates of that
>same individual in the second dataset, creating
>1 dataset with 2 columns of landmark
>coordinates? In that case there should be at
>least 1 landmark that is present in both dataset
>as a link, right? The Adams (1999) paper
>mentions including also the ratio of size of the
>2 datasets i.e. centroid size when using the
>separate subset method. How should you include this in the compiled xls?
>Â
>Thanks in advance.
>Â
>Kind regards,
>Pepijn Kamminga
>Â
>PhD candidate
>Naturalis Biodiversity Center, Leiden, The Netherlands
>Leiden University, The Netherlands
>email: <mailto:[email protected]>[email protected]
>
>2013/3/27
><<mailto:[email protected]>[email protected]>
>
>----- Forwarded message from andrea cardini
><<mailto:[email protected]>[email protected]> -----
>
>Â Â Â Date: Tue, 26 Mar 2013 07:12:52 -0400
>Â Â Â From: andrea cardini
><<mailto:[email protected]>[email protected]>
>Â Â Â Reply-To: andrea cardini
><<mailto:[email protected]>[email protected]>
>Â Â Â Subject: Re: combining two data set
>Â Â Â To: <mailto:[email protected]>[email protected]
>
>Dear Wendy,
>There's a good chance that you may find the
>answer to your question in this paper:
><http://www.public.iastate.edu/~dcadams/PDFPubs/1999-Adams-EER.pdf>http://www.public.iastate.edu/~dcadams/PDFPubs/1999-Adams-EER.pdf
>
>I have the suspicion that in your case appending
>the shape variables might be the option. Easy to
>do in xls etc. The appended matrices will
>probably have to be analysed in another software
>(PAST, R or any commercial stats soft.). The
>visualization will have to be done for each
>separate structure, however, if you use this method.
>
>Linking in MorphoJ, Â as far as I know, has a
>different meaning and it is in the context of the
>PLS or similar analyses (for instance, if I aim
>to measure the covariation between the mandible
>and cranium, that might be the analysis). If
>that's you purpose, it's even easier (doable in
>MorphoJ, TPSPLS and probably PAST).
>
>Cheers
>
>Andrea
>
>At 06:23 26/03/2013, you wrote:
>
> >----- Forwarded message from Wendy Den Boer
> <<mailto:[email protected]>[email protected]> -----
> >
> > Â Â Â Date: Mon, 25 Mar 2013 07:31:42 -0400
> > Â Â Â From: Wendy Den Boer
> <<mailto:[email protected]>[email protected]>
> > Â Â Â Reply-To: Wendy Den Boer
> <<mailto:[email protected]>[email protected]>
> > Â Â Â Subject: combining two data set
> > Â Â Â To: <mailto:[email protected]>[email protected]
> >
> >Hi,
> >
> >I have two data sets, one of several specimens with their landmark
> >values in lateral view, and then the exact same specimens but then
> >with their landmark values in dorsal view. These landmarks were
> >obtained in tpsDIG.
> >
> >I don't want to compare the dorsal and lateral views of these
> >specimens to one another but I want to combine them. Hence, how do I
> >manipulate tpsRELW so that it doesn't treat the dorsal and ventral as
> >separate individuals, but uses both the ventral and dorsal landmark
> >values to produce one individual output for each specimen? Hope this
> >makes sense. If it isn't possible to do this in tpsRELW would I have
> >to get the mean values of the x coordinates of the dorsal and lateral
> >views, and the mean values of the y coordinates of the dorsal and
> >lateral view of each specimen in order to do this?
> >If anyone knows how to do this in MorphoJ and Past that would be great
> >as well. Â I know there is a link data-sets option in MorphoJ that
> >requires the identifiers in the two data-sets to be the same or that
> >they match. However, when I try to link the two data-sets nothing
> >seems to happen, unless I'm missing the obvious somehow. Sorry, if I am.
> >
> >Many Thanks,
> >
> >Wendy den Boe
> >
> >----- End forwarded message -----
>
>Dr. Andrea Cardini
>Researcher in Animal Biology
>Dipartimento di Scienze Chimiche e Geologiche,
>Università di Modena e Reggio Emilia, l.go S. Eufemia 19, 41121 Modena, Italy
>
>Honorary Fellow
>Centre for Anatomical and Human Sciences
>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:
><mailto:[email protected]>[email protected],
><mailto:[email protected]>[email protected]
>
>Webpage:
><http://sites.google.com/site/hymsfme/drandreacardini>http://sites.google.com/site/hymsfme/drandreacardini
>Datasets:
><http://ads.ahds.ac.uk/catalogue/archive/cerco_lt_2007/overview.cfm#metadata>http://ads.ahds.ac.uk/catalogue/archive/cerco_lt_2007/overview.cfm#metadata
>Editorial board for:
>Zoomorphology:
><http://www.springer.com/life+sciences/animal+sciences/journal/435>http://www.springer.com/life+sciences/animal+sciences/journal/435
>Journal of Zoological Systematics and
>Evolutionary Research:
><http://www.wiley.com/bw/journal.asp?ref=0947-5745&site=1>http://www.wiley.com/bw/journal.asp?ref=0947-5745&site=1
>Hystrix, the Italian Journal of
>Mammalogy:
><http://www.italian-journal-of-mammalogy.it/>http://www.italian-journal-of-mammalogy.it/
>
>----- End forwarded message -----
>
>
>
>
>
>----- End forwarded message -----
>
>
Dr. Andrea Cardini
Researcher in Animal Biology
Dipartimento di Scienze Chimiche e Geologiche,
Università di Modena e Reggio Emilia, l.go S. Eufemia 19, 41121 Modena, Italy
Honorary Fellow
Centre for Anatomical and Human Sciences
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]
Webpage: http://sites.google.com/site/hymsfme/drandreacardini
Datasets:
http://ads.ahds.ac.uk/catalogue/archive/cerco_lt_2007/overview.cfm#metadata
Editorial board for:
Zoomorphology:
http://www.springer.com/life+sciences/animal+sciences/journal/435
Journal of Zoological Systematics and
Evolutionary Research: http://www.wiley.com/bw/journal.asp?ref=0947-5745&site=1
Hystrix, the Italian Journal of
Mammalogy: http://www.italian-journal-of-mammalogy.it/
----- End forwarded message -----
