Unwanted landmark mirroring after GPA

[agne89]

Hi,

I am doing a 3D analysis of symmetry in dolphin skulls using landVR.

To do this, I need to create a new shape array where half of the skull landmarks (single points only) are mirrored to the otehr half of the skull, to be comprared with the real data I manually placed on both halves of the skulls.

The mirrored dataset construted with the "mirrorfill" function in paleomorph looks ok, but after GPA in geomorph the landmark order in the aligned coordinates has changed, with the two halves beign flipped (waht wa son the left was on the right and vice versa).

This of course makes using landVR impossible. The GPA fucntion in Morpho (procSym) doesn't have this problem.

Is there a setting I can change to avoid this? 

I attach some code and pictures to demonstrate the problem.

Thank you, best wishes

Agnese

#Mirror left on right

shape_array_LM_mirror <- mirrorfill(A = mirror_array,  l1 = as.integer(midline), l2 = bilat.landmarks) # the NA rows are now filled so that the numbers are the same on both

#Plot mirrored landmarks

mirror_plot <- c(#left red

                 spheres3d(shape_array_LM_mirror[landpairs_L,,23], col = mypalette_paired[6], radius = 1.5),

                 #right blue

                 spheres3d(shape_array_LM_mirror[landpairs_R,,23], col = mypalette_paired[2], radius = 1.5),

                 #midline dark gray

                 spheres3d(shape_array_LM_mirror[midline,,23], col = "gray10", radius = 1.5))

 

#GPA mirrored landmarks

#Geomorph - problem reflection landmarks

gpa_mirror <- gpagen(shape_array_LM_mirror)

#Check LMs

mirror_plot_gpa <- c(#left red

                     spheres3d(gpa_mirror$coords [landpairs_L,,23], col = mypalette_paired[6], radius = 0.003),

                     #right blue

                     spheres3d(gpa_mirror$coords[landpairs_R,,23], col = mypalette_paired[2], radius = 0.003),

                     #midline dark gray

                     spheres3d(coords_mirror[midline,,23], col = "gray10", radius = 0.003))

#Morpho - ok

gpa_mirror_morpho <- procSym(shape_array_LM_mirror)

#Check LMs

mirror_plot_gpa_morpho <- c(#left red

  spheres3d(gpa_mirror_morpho$rotated[landpairs_L,,23], col = mypalette_paired[6], radius = 0.003),

  #right blue

  spheres3d(gpa_mirror_morpho$rotated[landpairs_R,,23], col = mypalette_paired[2], radius = 0.003),

  #midline dark gray

  spheres3d(gpa_mirror_morpho$rotated[midline,,23], col = "gray10", radius = 0.003))

[Mike Collyer]

Agnese,

It’s a good bet that this is the arbitrary flipping of PC axes, as you aligned configurations to principal axes (PCs) in both the geomorph and Morpho cases, by choosing the default options.  (Small procedural note: Morpho uses the PCs of the first specimen; geomorph uses the PCs of the reference [mean] configuration.)

You can tell by the black points that the vertical axis (PC 2) is flipped rather than left becoming right and right becoming left.

This eigen-analysis artifact poses no consequence for your data.  If it is aesthetically troubling, you can use (geomorph) rotate.coords to flip the axes as you like.

Mike

On Nov 20, 2021, at 11:46 AM, agne89 <agnese....@hotmail.it> wrote:

Hi,

I am doing a 3D analysis of symmetry in dolphin skulls using landVR.

To do this, I need to create a new shape array where half of the skull landmarks (single points only) are mirrored to the otehr half of the skull, to be comprared with the real data I manually placed on both halves of the skulls.

The mirrored dataset construted with the "mirrorfill" function in paleomorph looks ok, but after GPA in geomorph the landmark order in the aligned coordinates has changed, with the two halves beign flipped (waht wa son the left was on the right and vice versa).

This of course makes using landVR impossible. The GPA fucntion in Morpho (procSym) doesn't have this problem.

Is there a setting I can change to avoid this? 

I attach some code and pictures to demonstrate the problem.

Thank you, best wishes

Agnese

#Mirror left on right

shape_array_LM_mirror <- mirrorfill(A = mirror_array,  l1 = as.integer(midline), l2 = bilat.landmarks) # the NA rows are now filled so that the numbers are the same on both

#Plot mirrored landmarks

mirror_plot <- c(#left red

                 spheres3d(shape_array_LM_mirror[landpairs_L,,23], col = mypalette_paired[6], radius = 1.5),

                 #right blue

                 spheres3d(shape_array_LM_mirror[landpairs_R,,23], col = mypalette_paired[2], radius = 1.5),

                 #midline dark gray

                 spheres3d(shape_array_LM_mirror[midline,,23], col = "gray10", radius = 1.5))

<mirror_LMs.png> 

#GPA mirrored landmarks

#Geomorph - problem reflection landmarks

gpa_mirror <- gpagen(shape_array_LM_mirror)

#Check LMs

mirror_plot_gpa <- c(#left red

                     spheres3d(gpa_mirror$coords [landpairs_L,,23], col = mypalette_paired[6], radius = 0.003),

                     #right blue

                     spheres3d(gpa_mirror$coords[landpairs_R,,23], col = mypalette_paired[2], radius = 0.003),

                     #midline dark gray

                     spheres3d(coords_mirror[midline,,23], col = "gray10", radius = 0.003))

<mirror_LMs_gpa.png>

#Morpho - ok

gpa_mirror_morpho <- procSym(shape_array_LM_mirror)

#Check LMs

mirror_plot_gpa_morpho <- c(#left red

  spheres3d(gpa_mirror_morpho$rotated[landpairs_L,,23], col = mypalette_paired[6], radius = 0.003),

  #right blue

  spheres3d(gpa_mirror_morpho$rotated[landpairs_R,,23], col = mypalette_paired[2], radius = 0.003),

  #midline dark gray

  spheres3d(gpa_mirror_morpho$rotated[midline,,23], col = "gray10", radius = 0.003))

<mirror_LMs_morpho.png>

--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/3d7b95c6-5a20-477f-9770-67b2a06a0539n%40googlegroups.com.
<mirror_LMs.png><mirror_LMs_gpa.png><mirror_LMs_morpho.png>

[agne89]

Hi Michael,

thanks a lot for your answer and explanation.

I think though it is not just a plotting problem, or at least it's a problem when trying to use landVR. This package measures the differences between 2 sets of landmarks, one mirrored and one manually placed. They are aligned separately.

As the axes are flipped in the mirror one, this creates artifically inflated differences with the manually placed data that don't get flipped (likely because of the strong differences between the 2 sides of the skull).

 

I attach the landVR visualization of procustes distances between landmarks of the same set of specimens. The first one was done with the flipped dataset from geomorph, and the vectors are pretty extreme. The other one is the correct one done with the Morpho-aligned coordinates.

Would apply rotate.coords fix this? I read on the help file that it only works on 2D landarmarks and I have 3D data.

Thank you, best wishes

Agnese

[Mike Collyer]

Agnese,

Which version of geomorph are you using?  There was an issue like this that resulted from a bug that we fixed many versions ago.  I am trying to understand if this could be the same issue.

Thanks!

Mike

On Nov 20, 2021, at 1:54 PM, agne89 <agnese....@hotmail.it> wrote:

Hi Michael,

thanks a lot for your answer and explanation.

I think though it is not just a plotting problem, or at least it's a problem when trying to use landVR. This package measures the differences between 2 sets of landmarks, one mirrored and one manually placed. They are aligned separately.

As the axes are flipped in the mirror one, this creates artifically inflated differences with the manually placed data that don't get flipped (likely because of the strong differences between the 2 sides of the skull).

 

I attach the landVR visualization of procustes distances between landmarks of the same set of specimens. The first one was done with the flipped dataset from geomorph, and the vectors are pretty extreme. The other one is the correct one done with the Morpho-aligned coordinates.

<image.png><radii_diff_phoc1.png>

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/daa848e5-926c-4e00-ba46-b242d38571edn%40googlegroups.com.
<image.png><radii_diff_phoc1.png>

[Agnese Lanzetti]

Hi Mike,

 

I am using 4.0.1 which I think it’s the most up to date version on CRAN.

 

Thank you, best

Agnese

--
You received this message because you are subscribed to a topic in the Google Groups "geomorph R package" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/geomorph-r-package/JIlV_9ZbrGU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to geomorph-r-pack...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/7198310E-6817-40CB-B4CA-AAD197FC29A5%40gmail.com.

 

[Mike Collyer]

OK.  Unfortunately, I do not have any familiarity with landVR, paleomorph, or the function, mirrorfill.  Maybe another user in this group has some experience using these functions along with geomorph and might have some ideas.  

Good luck!

Mike

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/BYAPR01MB4791EE2B8E36A13ED5A688E5939D9%40BYAPR01MB4791.prod.exchangelabs.com.

[Mike Collyer]

One other thing came to mind that maybe could be an issue.  In your analyses, is it possible that you have configurations that are reflections?  For example, are you mirroring the right side on some and the left side on others?  The procSym function has an option to reflect configurations (default is TRUE), but gpagen does not.  

This is just a guess, as it is not clear without knowing how the function works.

Best,

Mike

[Agnese Lanzetti]

Hi Mike, 

No it's always the left side mirrored to the right. 

For now I'll align the landmarks with Morpho for landVR to be safe. 

Thank you, best regards 

Agnese 

On Nov 20, 2021 at 8:22 PM, Mike Collyer <mlco...@gmail.com> wrote:

One other thing came to mind that maybe could be an issue.  In your analyses, is it possible that you have configurations that are reflections?  For example, are you mirroring the right side on some and the left side on others?  The procSym function has an option to reflect configurations (default is TRUE), but gpagen does not.  

This is just a guess, as it is not clear without knowing how the function works.

Best,

Mike

On Nov 20, 2021, at 2:31 PM, Mike Collyer <mlco...@gmail.com> wrote:

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/1B3E1140-3039-4C2C-81CE-1F10E5BE0DBA%40gmail.com.

[agne89]

Hi Mike,

I just have a quick follow up question, as I'm trying to somewhat replicate the results of the bilat.symmetry function in landVR. 

When I plot the bilat.symmetry obejct, what shapes are the gray and black dots based on for directional asymmetry?

They are clearly not the same between the directional and fluctuating asymmetry plot, but I don't know what the ones in the DA plot are.

I attach an example.

Thank you, best

Agnese

[Adams, Dean [EEOB]]

Agnese,

 

The plots from bilat.symmetry are shunted through plotRefToTarget.

 

In that function, the gray dots are the data, and the black dots the overall reference. For bilat.symmetry plots, the gray dots show the directional and fluctuating components respectively in the L/R plot panels.

 

Dean

 

Dr. Dean C. Adams (he/him)

Distinguished Professor of Evolutionary Biology

Director of Graduate Education, EEB Program

Department of Ecology, Evolution, and Organismal Biology

Iowa State University

https://faculty.sites.iastate.edu/dcadams/

phone: 515-294-3834

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/7c445d68-99be-4ee8-8b42-bbeff1e7c13bn%40googlegroups.com.

[Agnese Lanzetti]

Hi Dean, 

Thank you for your answer. 

So the gray dots are the DA component for each side? As the function produces two sets of mean shape coordinates for DA for each side. 

The black dots instead are the mean symmetric shape I assume? 

Thank you, best regards 

Agnese 

On Nov 23, 2021 at 3:14 PM, Adams, Dean [EEOB] <dca...@iastate.edu> wrote:

Agnese,

 

The plots from bilat.symmetry are shunted through plotRefToTarget.

 

In that function, the gray dots are the data, and the black dots the overall reference. For bilat.symmetry plots, the gray dots show the directional and fluctuating components respectively in the L/R plot panels.

 

Dean

 

Dr. Dean C. Adams (he/him)

Distinguished Professor of Evolutionary Biology

Director of Graduate Education, EEB Program

Department of Ecology, Evolution, and Organismal Biology

Iowa State University

https://faculty.sites.iastate.edu/dcadams/

phone: 515-294-3834

 

From: geomorph-...@googlegroups.com <geomorph-...@googlegroups.com> On Behalf Of agne89
Sent: Tuesday, November 23, 2021 3:59 AM

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/7c445d68-99be-4ee8-8b42-bbeff1e7c13bn%40googlegroups.com.

--
You received this message because you are subscribed to a topic in the Google Groups "geomorph R package" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/geomorph-r-package/JIlV_9ZbrGU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to geomorph-r-pack...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/CH2PR04MB66171BD1919CA90C85783612A2609%40CH2PR04MB6617.namprd04.prod.outlook.com.

[Adams, Dean [EEOB]]

Yes. See examples in the help file

 

Dr. Dean C. Adams (he/him)

Distinguished Professor of Evolutionary Biology

Director of Graduate Education, EEB Program

Department of Ecology, Evolution, and Organismal Biology

Iowa State University

https://faculty.sites.iastate.edu/dcadams/

phone: 515-294-3834

 

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/BYAPR01MB47911B170F1E81FF9D30D46E93609%40BYAPR01MB4791.prod.exchangelabs.com.

[Agnese Lanzetti]

Hi Dean, 

Thank you for clarifying this. 

Best, 

Agnese 

On Nov 23, 2021 at 5:18 PM, Adams, Dean [EEOB] <dca...@iastate.edu> wrote:

Yes. See examples in the help file

 

Dr. Dean C. Adams (he/him)

Distinguished Professor of Evolutionary Biology

Director of Graduate Education, EEB Program

Department of Ecology, Evolution, and Organismal Biology

Iowa State University

https://faculty.sites.iastate.edu/dcadams/

phone: 515-294-3834

 

From: geomorph-...@googlegroups.com <geomorph-...@googlegroups.com> On Behalf Of Agnese Lanzetti

Sent: Tuesday, November 23, 2021 11:02 AM

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/BYAPR01MB47911B170F1E81FF9D30D46E93609%40BYAPR01MB4791.prod.exchangelabs.com.

--
You received this message because you are subscribed to a topic in the Google Groups "geomorph R package" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/geomorph-r-package/JIlV_9ZbrGU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to geomorph-r-pack...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/CH2PR04MB6617EEAD08C61B7C00AF1BE8A2609%40CH2PR04MB6617.namprd04.prod.outlook.com.