Anderson,

I want to add another point the others have not yet mentioned.

One very simple explanation could be that you are seeing parallax effects,
although it is hard to be sure without knowing which group represents which
magnification or where the points are on the skull.  At a minimum, the
camera will see less of the surface if it is closer to the surface – see
(A) in the attached figure for an extreme example (the reason is when the
camera is closer to the object, a ray of light traveling to the camera
aperture will be tangent to the surface for points closer to the center of
the field of view).  For related reasons, relief features will look
different due to both the camera’s distance from the object (B). You can
also easily show that this effect will vary with the distance of the
feature from the center of the field of view.

For these reasons, it is important to photograph larger objects from a
proportionately greater distance.  It is also a good idea to have the
region centered in the field of view, although you will still have to
consider whether the changes in the position of a landmark is due to
displacement of that feature, change in its relative height, or both.

On Fri, Jan 12, 2018 at 2:04 AM, Anderson Feijo <andefe...@gmail.com> wrote:

> Dear Dean and Sonja,
>
> Thank you for your suggestions. I am exploring all these potential bias
> before starting my study. I have learned a lot through all comments and
> literature suggestions. I deeply appreciated your help.
>
> Best,
>
> Anderson
>
> On Tue, Jan 9, 2018 at 3:30 AM, Adams, Dean [EEOBS] <dcad...@iastate.edu>
> wrote:
>
>> Anderson,
>>
>>
>>
>> I don’t think you appreciated the importance of the Murat’s comments on
>> your earlier post on this same topic.
>>
>>
>>
>> In theory, there is no problem combining objects digitized at different
>> magnifications, or even digitized by different researchers. However, before
>> doing so one must carefully investigate for possible systematic biases in
>> digitizing, so they may be reduced to the greatest extent possible. If
>> there is some consistent bias in how objects are digitized in one ‘group’
>> relative to the other, this will permeate into perceived differences in
>> shape that may not exist. A common example with older digitizing tablets
>> would be differences in digitizing due to the handedness of the person
>> digitizing. Right-handed and left-handed individuals hold the stylus
>> differently which can result in consistent perceived shape differences of
>> the same objects once digitized.
>>
>>
>>
>> Whether or not you have such an issue with your two magnifications is
>> unclear. However, it is impossible to evaluate this without additional
>> replication. Again, as Murat suggested, try digitizing each object multiple
>> times at each magnification. Then one could obtain estimates of the
>> variation in digitizing at the same magnification versus across
>> magnifications to begin to discern whether the between-magnification
>> variation is greater than one might expect. If it is, then one must dig
>> deeper to determine why.
>>
>>
>>
>> I would recommend sorting all of this out before embarking on your
>> empirical study. Otherwise, interpreting patterns in the final dataset
>> becomes challenging to say the least.
>>
>>
>> Best,
>>
>>
>>
>> Dean
>>
>>
>>
>> Dr. Dean C. Adams
>>
>> Professor
>>
>> Department of Ecology, Evolution, and Organismal Biology
>>
>>        Department of Statistics
>>
>> Iowa State University
>>
>> www.public.iastate.edu/~dcadams/
>>
>> phone: 515-294-3834 <(515)%20294-3834>
>>
>>
>>
>> *From:* Anderson Feijo [mailto:andefe...@gmail.com]
>> *Sent:* Friday, December 29, 2017 3:19 AM
>> *To:* MORPHMET <morphmet@morphometrics.org>
>> *Subject:* [MORPHMET] Doubt about scalling photos
>>
>>
>>
>> Hi everyone,
>>
>>
>>
>> I am starting a new project using GM which I will work with groups with
>> different sizes (e.g., rodents and small carnivores). I would like to find
>> a way to use the whole dataset in the analyses, instead of perform set of
>> analyses for each sized group. So, I did a test using one skull and place
>> the camera in two different distances to the object (~15 cm and ~30 cm). My
>> expectation was after scaling (using tpsDig) I wouldn´t have any meaningful
>> difference. But I got two clear groups that were statistically different.
>> So, my question is how can I combine 2D landmarks based on photos taken
>> from different distances of the camera to the object. I have attached here
>> the tps file (10 copies of the same skull, five at ~15cm and five at
>> ~30cm). I would be very grateful for any suggestion.
>>
>>
>>
>> All the best and Happy 2018!
>>
>>
>>
>> Anderson
>>
>> --
>> 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.
>>
>
>
>
> --
> _____________________________________________
>
> Dr. Anderson Feijó
>
> Key Laboratory of Zoological Systematics and Evolution
> Institute of Zoology, Chinese Academy of Science
> Beichen West Road, Chaoyang District, 100101
> Beijing, China
>
> Curriculum: *Lattes
> <http://lattes.cnpq.br/9406413385468571>; ResearchGate
> <https://www.researchgate.net/profile/Anderson_Feijo>*
>
> --
> 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.
>



-- 
Donald L Swiderski
University of Michigan
ph.(734) 763-9613
e-mail: dlswi...@umich.edu

-- 
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.

Reply via email to