Re: [MORPHMET] Interpreting PC scores graph in MorphoJ when the first two PC's are difference due to operator input
Hi Hadas, Thank you for the reply. The test has been conducted previously before finalising the landmark points for other users. I suspect the raw bone data itself is unsatisfactory. The .obj files generated from PACS>MITK>MeshLab>IDAVLandmark to have problems. Kind regards, Helmi On Tuesday, October 11, 2016 at 4:50:12 AM UTC+8, Hadas L. wrote: > > Hi, > > I would consider doing a short reliability test on five specimens for each > individual for the single point landmarks and the curves. This shouldn't > take long and will give you an idea as to whether the landmarks are > reliable or not. Ideally this should be done before adding the landmarks to > the data set. > > Sincerely, > Hadas > > On Sun, Oct 9, 2016 at 5:56 AM, Helmi Hadi > wrote: > >> Dear Morphometricians, >> >> I have a weird problem and I was hoping someone could help me on this. >> >> I have the same CT bone (n=400) with 14 landmarks (2 Single point and 12 >> sliding) larndmarked in IDAV Landmark by three different individuals >> following a figure key table. Each individual landmarked 100+ different >> bones. The figure key table has the shape of the bone, with the landmark >> locations and the order of landmarking. All users were briefed by a single >> person (user1) on how to extract the bone and landmark. All three >> individuals were present throughout the entire procedure but the bone >> segmentation and landmarking process was conducted individually. >> >> When I combine all the data, I noticed that PC1 and PC2 graph has two >> clusters. The effect is about 55% (as detailed in the eigenvalues below) >> After further checking the classifiers in MorphoJ, the source of the >> clustering is one person (user1) landmarked it is slightly differently >> compared to the other two. I have checked the outliers tab and no glaring >> outliers exists. As the sample size is big, the curve seem to be quite >> normal. >> >> Eigenvalues % Variance Cumulative % >> 1. 0.01603351 55.674 55.674 >> 2. 0.00496222 17.230 72.904 >> 3. 0.00224594 7.799 80.703 >> 4. 0.00121085 4.204 84.907 >> 5. 0.00076858 2.669 87.576 >> ... >> >> How to interpret the results for this kind of data? Things which come to >> my mind are: >> >> 1. Maybe the bones for user1 are different compared to the other two >> users. >> Or >> 2. User1 thinks the landmark location slightly different compared to >> others. >> Or >> 3. User2 and 3 could not locate the landmark locations of user1. >> Or >> 4. The landmarks selected are unreliable. >> >> Ideally I would need few other people relandmark the entire set, but it >> is not possible to do it now. Can anyone help shed some light on what is >> the probable cause? Thank you. >> >> >> Kind regards, >> >> Helmi Hadi, PhD >> School of Health Sciences, Universiti Sains Malaysia, >> 16150 Kubang Kerian, Kelantan, MALAYSIA >> >> >> -- >> 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+u...@morphometrics.org . >> > > -- 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.
Re: [MORPHMET] Interpreting PC scores graph in MorphoJ when the first two PC's are difference due to operator input
Dear Carmelo, Thank you. I will try to look at the at that method. Thank you also for giving the reference material. The landmarks chosen has been tested to be replicable based on prior testing before deploying to other users. The first person briefed and thought all the other users the landmarking procedure and the segmentation and cleaning of the bone in PACS (a CT database). I will have a look at raw ct files first and try to get it landmarked. Maybe some people see the semi-landmarks slightly differently. Kind regards, Helmi On Monday, October 10, 2016 at 8:28:32 AM UTC+8, Carmelo Fruciano wrote: > > Dear Helmi, > I won't answer you in detail on your specific case as I didn't fully > understand the problem/design. > > Ideally, if you have at least a subset of the same bones being > digitized by all the users/operators, you can check for your > hypothesis 1. > > In some cases it might be possible to "remove" measurement error from > your data using the method presented in > > Valentin AE, Penin X, Chanut JP, Sévigny JM, Rohlf FJ (2008) Arching > effect on fish body shape in geometric morphometric studies. Journal of > Fish > Biology 73:623–638. doi:10./j.1095-8649.2008.01961.x > > That method and other approaches/issues are also discussed in my > recent review on the topic of measurement error > > Fruciano C (2016) Measurement error in geometric morphometrics. > Development Genes and Evolution, 226(3):139–158. doi: > 10.1007/s00427-016-0537-4. > > which might be useful to you to understand what's possible to do in > your specific case. > > I hope this helps, > Carmelo > > > > > > Helmi Hadi > ha scritto: > > > Dear Morphometricians, > > > > I have a weird problem and I was hoping someone could help me on this. > > > > I have the same CT bone (n=400) with 14 landmarks (2 Single point and 12 > > sliding) larndmarked in IDAV Landmark by three different individuals > > following a figure key table. Each individual landmarked 100+ different > > bones. The figure key table has the shape of the bone, with the landmark > > locations and the order of landmarking. All users were briefed by a > single > > person (user1) on how to extract the bone and landmark. All three > > individuals were present throughout the entire procedure but the bone > > segmentation and landmarking process was conducted individually. > > > > When I combine all the data, I noticed that PC1 and PC2 graph has two > > clusters. The effect is about 55% (as detailed in the eigenvalues below) > > After further checking the classifiers in MorphoJ, the source of the > > clustering is one person (user1) landmarked it is slightly differently > > compared to the other two. I have checked the outliers tab and no > glaring > > outliers exists. As the sample size is big, the curve seem to be quite > > normal. > > > > Eigenvalues % Variance Cumulative % > > 1. 0.01603351 55.674 55.674 > > 2. 0.00496222 17.230 72.904 > > 3. 0.00224594 7.799 80.703 > > 4. 0.00121085 4.204 84.907 > > 5. 0.00076858 2.669 87.576 > > ... > > > > How to interpret the results for this kind of data? Things which come to > my > > mind are: > > > > 1. Maybe the bones for user1 are different compared to the other two > users. > > Or > > 2. User1 thinks the landmark location slightly different compared to > others. > > Or > > 3. User2 and 3 could not locate the landmark locations of user1. > > Or > > 4. The landmarks selected are unreliable. > > > > Ideally I would need few other people relandmark the entire set, but it > is > > not possible to do it now. Can anyone help shed some light on what is > the > > probable cause? Thank you. > > > > > > Kind regards, > > > > Helmi Hadi, PhD > > School of Health Sciences, Universiti Sains Malaysia, > > 16150 Kubang Kerian, Kelantan, MALAYSIA > > > > > > -- > > 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+u...@morphometrics.org . > > > > > > -- > Carmelo Fruciano > Postdoctoral Fellow - Queensland University of Technology - Brisbane, > Australia > Honorary Fellow - University of Catania - Catania, Italy > e-mail c.fru...@unict.it > http://www.fruciano.it/research/ > > -- 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.
Re: [MORPHMET] Interpreting PC scores graph in MorphoJ when the first two PC's are difference due to operator input
Hi, I would consider doing a short reliability test on five specimens for each individual for the single point landmarks and the curves. This shouldn't take long and will give you an idea as to whether the landmarks are reliable or not. Ideally this should be done before adding the landmarks to the data set. Sincerely, Hadas On Sun, Oct 9, 2016 at 5:56 AM, Helmi Hadi wrote: > Dear Morphometricians, > > I have a weird problem and I was hoping someone could help me on this. > > I have the same CT bone (n=400) with 14 landmarks (2 Single point and 12 > sliding) larndmarked in IDAV Landmark by three different individuals > following a figure key table. Each individual landmarked 100+ different > bones. The figure key table has the shape of the bone, with the landmark > locations and the order of landmarking. All users were briefed by a single > person (user1) on how to extract the bone and landmark. All three > individuals were present throughout the entire procedure but the bone > segmentation and landmarking process was conducted individually. > > When I combine all the data, I noticed that PC1 and PC2 graph has two > clusters. The effect is about 55% (as detailed in the eigenvalues below) > After further checking the classifiers in MorphoJ, the source of the > clustering is one person (user1) landmarked it is slightly differently > compared to the other two. I have checked the outliers tab and no glaring > outliers exists. As the sample size is big, the curve seem to be quite > normal. > > Eigenvalues % Variance Cumulative % > 1. 0.01603351 55.674 55.674 > 2. 0.00496222 17.230 72.904 > 3. 0.00224594 7.799 80.703 > 4. 0.00121085 4.204 84.907 > 5. 0.00076858 2.669 87.576 > ... > > How to interpret the results for this kind of data? Things which come to > my mind are: > > 1. Maybe the bones for user1 are different compared to the other two users. > Or > 2. User1 thinks the landmark location slightly different compared to > others. > Or > 3. User2 and 3 could not locate the landmark locations of user1. > Or > 4. The landmarks selected are unreliable. > > Ideally I would need few other people relandmark the entire set, but it is > not possible to do it now. Can anyone help shed some light on what is the > probable cause? Thank you. > > > Kind regards, > > Helmi Hadi, PhD > School of Health Sciences, Universiti Sains Malaysia, > 16150 Kubang Kerian, Kelantan, MALAYSIA > > > -- > 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. > -- 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.
Re: [MORPHMET] Interpreting PC scores graph in MorphoJ when the first two PC's are difference due to operator input
Dear Helmi, I won't answer you in detail on your specific case as I didn't fully understand the problem/design. Ideally, if you have at least a subset of the same bones being digitized by all the users/operators, you can check for your hypothesis 1. In some cases it might be possible to "remove" measurement error from your data using the method presented in Valentin AE, Penin X, Chanut JP, Sévigny JM, Rohlf FJ (2008) Arching effect on fish body shape in geometric morphometric studies. Journal of Fish Biology 73:623–638. doi:10./j.1095-8649.2008.01961.x That method and other approaches/issues are also discussed in my recent review on the topic of measurement error Fruciano C (2016) Measurement error in geometric morphometrics. Development Genes and Evolution, 226(3):139–158. doi: 10.1007/s00427-016-0537-4. which might be useful to you to understand what's possible to do in your specific case. I hope this helps, Carmelo Helmi Hadi ha scritto: Dear Morphometricians, I have a weird problem and I was hoping someone could help me on this. I have the same CT bone (n=400) with 14 landmarks (2 Single point and 12 sliding) larndmarked in IDAV Landmark by three different individuals following a figure key table. Each individual landmarked 100+ different bones. The figure key table has the shape of the bone, with the landmark locations and the order of landmarking. All users were briefed by a single person (user1) on how to extract the bone and landmark. All three individuals were present throughout the entire procedure but the bone segmentation and landmarking process was conducted individually. When I combine all the data, I noticed that PC1 and PC2 graph has two clusters. The effect is about 55% (as detailed in the eigenvalues below) After further checking the classifiers in MorphoJ, the source of the clustering is one person (user1) landmarked it is slightly differently compared to the other two. I have checked the outliers tab and no glaring outliers exists. As the sample size is big, the curve seem to be quite normal. Eigenvalues % Variance Cumulative % 1. 0.01603351 55.674 55.674 2. 0.00496222 17.230 72.904 3. 0.00224594 7.799 80.703 4. 0.00121085 4.204 84.907 5. 0.00076858 2.669 87.576 ... How to interpret the results for this kind of data? Things which come to my mind are: 1. Maybe the bones for user1 are different compared to the other two users. Or 2. User1 thinks the landmark location slightly different compared to others. Or 3. User2 and 3 could not locate the landmark locations of user1. Or 4. The landmarks selected are unreliable. Ideally I would need few other people relandmark the entire set, but it is not possible to do it now. Can anyone help shed some light on what is the probable cause? Thank you. Kind regards, Helmi Hadi, PhD School of Health Sciences, Universiti Sains Malaysia, 16150 Kubang Kerian, Kelantan, MALAYSIA -- 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. -- Carmelo Fruciano Postdoctoral Fellow - Queensland University of Technology - Brisbane, Australia Honorary Fellow - University of Catania - Catania, Italy e-mail c.fruci...@unict.it http://www.fruciano.it/research/ -- 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.
[MORPHMET] Interpreting PC scores graph in MorphoJ when the first two PC's are difference due to operator input
Dear Morphometricians, I have a weird problem and I was hoping someone could help me on this. I have the same CT bone (n=400) with 14 landmarks (2 Single point and 12 sliding) larndmarked in IDAV Landmark by three different individuals following a figure key table. Each individual landmarked 100+ different bones. The figure key table has the shape of the bone, with the landmark locations and the order of landmarking. All users were briefed by a single person (user1) on how to extract the bone and landmark. All three individuals were present throughout the entire procedure but the bone segmentation and landmarking process was conducted individually. When I combine all the data, I noticed that PC1 and PC2 graph has two clusters. The effect is about 55% (as detailed in the eigenvalues below) After further checking the classifiers in MorphoJ, the source of the clustering is one person (user1) landmarked it is slightly differently compared to the other two. I have checked the outliers tab and no glaring outliers exists. As the sample size is big, the curve seem to be quite normal. Eigenvalues % Variance Cumulative % 1. 0.01603351 55.674 55.674 2. 0.00496222 17.230 72.904 3. 0.00224594 7.799 80.703 4. 0.00121085 4.204 84.907 5. 0.00076858 2.669 87.576 ... How to interpret the results for this kind of data? Things which come to my mind are: 1. Maybe the bones for user1 are different compared to the other two users. Or 2. User1 thinks the landmark location slightly different compared to others. Or 3. User2 and 3 could not locate the landmark locations of user1. Or 4. The landmarks selected are unreliable. Ideally I would need few other people relandmark the entire set, but it is not possible to do it now. Can anyone help shed some light on what is the probable cause? Thank you. Kind regards, Helmi Hadi, PhD School of Health Sciences, Universiti Sains Malaysia, 16150 Kubang Kerian, Kelantan, MALAYSIA -- 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.
