Re: [MORPHMET] old Morpheus et al. does truss, missing landmarks etc.

2016-11-14 Thread ahross
Dear All,
Morpheus et al. also has a data imputation feature.

Ann

Ann H. Ross, Ph.D., D-ABFA


> On Nov 14, 2016, at 3:42 AM, Thomas O'Mahoney  wrote:
> 
> Hi all,
> Both the R packages 'Morpho' and 'Geomorph' will allow you to estimate 
> missing landmarks using either thin plate splines or regression. It's about 3 
> lines of code, fully explained in each package, to import your data and then 
> estimate the missing data. Both packages accept a variety of file formats, 
> and can also export a variety as well, so you can do the remaining analysis 
> in MorphoJ if you prefer.
> Kind regards,
> Tom O'Mahoney 
> 
>> On Mon, 14 Nov 2016 at 07:34, alcardini  wrote:
>> Not exactly: another user-friendly option is the 'old Morpheus et
>> al.', which estimates missing landmarks (with coordinates, for
>> instance, coded in NTS as - - -, for each 3D landmark)
>> using mean substitution, regression, tps or mirror reflection. The
>> last option is a bit more complicated, I believe, and I never tried
>> it, but the other 3 are easy to select: import the NTS file and first
>> use the command SET FULLPRECISION ON; then try the command LIST PMISS
>> OPTIONS, and then SET PMISS IMPUTATION REGRESSION (for instance);
>> finally use SUPER GPA followed by PMISS IMPUTE; if you restore scale
>> (SUPER RESTORE SCALE), you'll get size back in the data (and also
>> estimated for specimens with missing landmarks).
>> You just need to re-export the nts file and remove, in the first line,
>> the code (1 -) for the missing landmarks (replace with simply 0):
>> check any description of NTS in the TPS Series and you'll see what I
>> mean.
>> 
>> In a couple of my old papers (the one with Thorington on marmots
>> ontogeny, 2007, and the one with Elton, 2008, BJLS on guenon skulls -
>> pdfs in my webpage) there might be something on this and how we
>> estimated the accuracy of missing landmarks estimates. Much more on
>> this topic is in several papers by Gunz and the others from the
>> Viennese school (including, if I am correct, one of their two papers
>> in the Hystrix, 2013, open access special issue).
>> 
>> 
>> As someone was also interested in the truss method (implemented in the
>> old Morpheus et al. as well), I've uploaded the software here:
>> https://www.dropbox.com/sh/ee5gu0qiqe6hz0v/AABLZkLs98uD8aXr79VaxbWma?dl=0
>> That should be the  latest version I have, which Dennis gave us in
>> Vienna 10 years ago. To be honest, I suspect it's still on the web, in
>> its official page, but haven't check the link for a while.
>> 
>> Dennis can help more with the 'secret commands', and check if I made
>> mistakes above.
>> There should be a few other people on the list with experience on
>> this, and certainly someone who has used the mirror reflection option
>> (on which, I can't help).
>> 
>> 
>> Good luck. Cheers
>> 
>> Andrea
>> 
>> 
>> On 13/11/2016, Murat Maga  wrote:
>> > MorphoJ handles missing landmarks, but I don't think it has a function to
>> > estimate them. SO without an estimation of your landmark position, your
>> > options are either to drop the sample from the analysis (if you want to
>> > retain full set of landmarks) or to remove the landmark from your analysis
>> > (to keep full set of individuals).
>> >
>> > You can also experiment with trying to estimate them based on reflection 
>> > (if
>> > it is a symmetrical structure) or based on bunch of reference samples. Both
>> > Morpho and gemorph in R have functions for estimating missing landmarks. 
>> > See
>> > if they give you reliable estimates, and then you can proceed with your
>> > analysis.
>> >
>> > Otherwise your options are limited to those two.
>> > M
>> >
>> >
>> > -Original Message-
>> > From: Rosa Perez [mailto:rmper...@ncsu.edu]
>> > Sent: Saturday, November 12, 2016 9:53 PM
>> > To: MORPHMET 
>> > Subject: [MORPHMET] Re: Problem with missing data using MorphoJ
>> >
>> > On Thursday, August 11, 2016 at 4:06:01 PM UTC-4, Jade Racine wrote:
>> >> Dear all,
>> >>
>> >>
>> >>
>> >> I digitized 108 craniofacial landmarks on human skulls from an
>> >> archaeological context using a MicroScribe G2X. I have a lot of random
>> >> missing values due to postmortem damage. I am currently trying to
>> >> analyze my data using MorphoJ. I followed the instructions from the
>> >> user’s guide, entering "-" in the data files for missing values. I
>> >> then combined the dorsal and ventral views using FileConverter. The files
>> >> I get can be read in MorphoJ.
>> >> However, nothing else works. I can’t visualize the landmarks or
>> >> perform any kind of analysis. If I try to find outliers, I get the
>> >> message "Finding outliers is not possible because only a single
>> >> observation or none at all is available". I get this message
>> >> regardless of whether I upload one file with all individuals in it or
>> >> multiple files with one individual 

Re: [MORPHMET] old Morpheus et al. does truss, missing landmarks etc.

2016-11-14 Thread Thomas O'Mahoney
Hi all,
Both the R packages 'Morpho' and 'Geomorph' will allow you to estimate
missing landmarks using either thin plate splines or regression. It's about
3 lines of code, fully explained in each package, to import your data and
then estimate the missing data. Both packages accept a variety of file
formats, and can also export a variety as well, so you can do the remaining
analysis in MorphoJ if you prefer.
Kind regards,
Tom O'Mahoney

On Mon, 14 Nov 2016 at 07:34, alcardini  wrote:

> Not exactly: another user-friendly option is the 'old Morpheus et
> al.', which estimates missing landmarks (with coordinates, for
> instance, coded in NTS as - - -, for each 3D landmark)
> using mean substitution, regression, tps or mirror reflection. The
> last option is a bit more complicated, I believe, and I never tried
> it, but the other 3 are easy to select: import the NTS file and first
> use the command SET FULLPRECISION ON; then try the command LIST PMISS
> OPTIONS, and then SET PMISS IMPUTATION REGRESSION (for instance);
> finally use SUPER GPA followed by PMISS IMPUTE; if you restore scale
> (SUPER RESTORE SCALE), you'll get size back in the data (and also
> estimated for specimens with missing landmarks).
> You just need to re-export the nts file and remove, in the first line,
> the code (1 -) for the missing landmarks (replace with simply 0):
> check any description of NTS in the TPS Series and you'll see what I
> mean.
>
> In a couple of my old papers (the one with Thorington on marmots
> ontogeny, 2007, and the one with Elton, 2008, BJLS on guenon skulls -
> pdfs in my webpage) there might be something on this and how we
> estimated the accuracy of missing landmarks estimates. Much more on
> this topic is in several papers by Gunz and the others from the
> Viennese school (including, if I am correct, one of their two papers
> in the Hystrix, 2013, open access special issue).
>
>
> As someone was also interested in the truss method (implemented in the
> old Morpheus et al. as well), I've uploaded the software here:
> https://www.dropbox.com/sh/ee5gu0qiqe6hz0v/AABLZkLs98uD8aXr79VaxbWma?dl=0
> That should be the  latest version I have, which Dennis gave us in
> Vienna 10 years ago. To be honest, I suspect it's still on the web, in
> its official page, but haven't check the link for a while.
>
> Dennis can help more with the 'secret commands', and check if I made
> mistakes above.
> There should be a few other people on the list with experience on
> this, and certainly someone who has used the mirror reflection option
> (on which, I can't help).
>
>
> Good luck. Cheers
>
> Andrea
>
>
> On 13/11/2016, Murat Maga  wrote:
> > MorphoJ handles missing landmarks, but I don't think it has a function to
> > estimate them. SO without an estimation of your landmark position, your
> > options are either to drop the sample from the analysis (if you want to
> > retain full set of landmarks) or to remove the landmark from your
> analysis
> > (to keep full set of individuals).
> >
> > You can also experiment with trying to estimate them based on reflection
> (if
> > it is a symmetrical structure) or based on bunch of reference samples.
> Both
> > Morpho and gemorph in R have functions for estimating missing landmarks.
> See
> > if they give you reliable estimates, and then you can proceed with your
> > analysis.
> >
> > Otherwise your options are limited to those two.
> > M
> >
> >
> > -Original Message-
> > From: Rosa Perez [mailto:rmper...@ncsu.edu]
> > Sent: Saturday, November 12, 2016 9:53 PM
> > To: MORPHMET 
> > Subject: [MORPHMET] Re: Problem with missing data using MorphoJ
> >
> > On Thursday, August 11, 2016 at 4:06:01 PM UTC-4, Jade Racine wrote:
> >> Dear all,
> >>
> >>
> >>
> >> I digitized 108 craniofacial landmarks on human skulls from an
> >> archaeological context using a MicroScribe G2X. I have a lot of random
> >> missing values due to postmortem damage. I am currently trying to
> >> analyze my data using MorphoJ. I followed the instructions from the
> >> user’s guide, entering "-" in the data files for missing values. I
> >> then combined the dorsal and ventral views using FileConverter. The
> files
> >> I get can be read in MorphoJ.
> >> However, nothing else works. I can’t visualize the landmarks or
> >> perform any kind of analysis. If I try to find outliers, I get the
> >> message "Finding outliers is not possible because only a single
> >> observation or none at all is available". I get this message
> >> regardless of whether I upload one file with all individuals in it or
> >> multiple files with one individual each. The problem seems to be with
> >> the "-" code because if I remove the missing landmarks from my
> >> data files, MorphoJ runs correctly. I tried to substitute "-" for
> >> "" or "-999". MorphoJ runs fine with those but generates odd
> >> results that do not look like the shape of a 

[MORPHMET] old Morpheus et al. does truss, missing landmarks etc.

2016-11-13 Thread alcardini
Not exactly: another user-friendly option is the 'old Morpheus et
al.', which estimates missing landmarks (with coordinates, for
instance, coded in NTS as - - -, for each 3D landmark)
using mean substitution, regression, tps or mirror reflection. The
last option is a bit more complicated, I believe, and I never tried
it, but the other 3 are easy to select: import the NTS file and first
use the command SET FULLPRECISION ON; then try the command LIST PMISS
OPTIONS, and then SET PMISS IMPUTATION REGRESSION (for instance);
finally use SUPER GPA followed by PMISS IMPUTE; if you restore scale
(SUPER RESTORE SCALE), you'll get size back in the data (and also
estimated for specimens with missing landmarks).
You just need to re-export the nts file and remove, in the first line,
the code (1 -) for the missing landmarks (replace with simply 0):
check any description of NTS in the TPS Series and you'll see what I
mean.

In a couple of my old papers (the one with Thorington on marmots
ontogeny, 2007, and the one with Elton, 2008, BJLS on guenon skulls -
pdfs in my webpage) there might be something on this and how we
estimated the accuracy of missing landmarks estimates. Much more on
this topic is in several papers by Gunz and the others from the
Viennese school (including, if I am correct, one of their two papers
in the Hystrix, 2013, open access special issue).


As someone was also interested in the truss method (implemented in the
old Morpheus et al. as well), I've uploaded the software here:
https://www.dropbox.com/sh/ee5gu0qiqe6hz0v/AABLZkLs98uD8aXr79VaxbWma?dl=0
That should be the  latest version I have, which Dennis gave us in
Vienna 10 years ago. To be honest, I suspect it's still on the web, in
its official page, but haven't check the link for a while.

Dennis can help more with the 'secret commands', and check if I made
mistakes above.
There should be a few other people on the list with experience on
this, and certainly someone who has used the mirror reflection option
(on which, I can't help).


Good luck. Cheers

Andrea


On 13/11/2016, Murat Maga  wrote:
> MorphoJ handles missing landmarks, but I don't think it has a function to
> estimate them. SO without an estimation of your landmark position, your
> options are either to drop the sample from the analysis (if you want to
> retain full set of landmarks) or to remove the landmark from your analysis
> (to keep full set of individuals).
>
> You can also experiment with trying to estimate them based on reflection (if
> it is a symmetrical structure) or based on bunch of reference samples. Both
> Morpho and gemorph in R have functions for estimating missing landmarks. See
> if they give you reliable estimates, and then you can proceed with your
> analysis.
>
> Otherwise your options are limited to those two.
> M
>
>
> -Original Message-
> From: Rosa Perez [mailto:rmper...@ncsu.edu]
> Sent: Saturday, November 12, 2016 9:53 PM
> To: MORPHMET 
> Subject: [MORPHMET] Re: Problem with missing data using MorphoJ
>
> On Thursday, August 11, 2016 at 4:06:01 PM UTC-4, Jade Racine wrote:
>> Dear all,
>>
>>
>>
>> I digitized 108 craniofacial landmarks on human skulls from an
>> archaeological context using a MicroScribe G2X. I have a lot of random
>> missing values due to postmortem damage. I am currently trying to
>> analyze my data using MorphoJ. I followed the instructions from the
>> user’s guide, entering "-" in the data files for missing values. I
>> then combined the dorsal and ventral views using FileConverter. The files
>> I get can be read in MorphoJ.
>> However, nothing else works. I can’t visualize the landmarks or
>> perform any kind of analysis. If I try to find outliers, I get the
>> message "Finding outliers is not possible because only a single
>> observation or none at all is available". I get this message
>> regardless of whether I upload one file with all individuals in it or
>> multiple files with one individual each. The problem seems to be with
>> the "-" code because if I remove the missing landmarks from my
>> data files, MorphoJ runs correctly. I tried to substitute "-" for
>> "" or "-999". MorphoJ runs fine with those but generates odd
>> results that do not look like the shape of a skull. I really want to
>> be able to analyze my dataset with missing values, otherwise I lose
>> too much landmarks and I can’t afford to remove individuals since my
>> sample is already small. Does anyone know what I am doing wrong or how I
>> can fix the problem? I attached an example of my data files.
>>
>>
>>
>>
>>
>> Any help would be greatly appreciated.
>>
>>
>> Thank you,
>>
>>
>>
>>
>>
>> Jade Racine
>>
>>
>> Master’s student
>>
>>
>> Department of Anthropology
>>
>>
>> University of Montreal, Canada
>
> Hi Jade,
>
> I am also a masters student attempting to use MorphoJ and am having a
> difficult time with the missing data. I collected 3D landmarks for a cranial
> asymmetry