Hi Andrea and Lydia just a word about "In the first case, however, CS is just rounding error (1.00001, 0.9999 and the like)". It's a while I don't use MorphoJ but I think that after GPA rotation with scaling at unit CS (thus "1" precisely) it projects data orthogonally [I dont think it does it stereo graphically] from the (hemi-) sphere [I suspect it does not scale at 1/cos(rho), i.e. Kendall's sphere sensu stricto] onto the tangent plane to the consensus; CS of shapes lying on the (hemi-) sphere is always actually 1, i.e. the sphere's radius by definition, but CS of projected data is the vector length from the (hemi-) sphere center and the point on the tangent plane; thus it is impossible to have CS values smaller than 1: all the values should exceed 1 by a small quantity but they should never be smaller nor exactly 1. Only the consensus will have 1 as it lies concomitantly on the sphere and on the tangent plane by definition. Not sure, however, about fine details in MorphoJ procedure.
All the best Paolo Il giorno gio 10 lug 2025 alle ore 09:33 andrea cardini <[email protected]> ha scritto: > > Dear Lydia, > I read quickly but the steps you wrote seem correct to me. Most of them > are also summarized in the guidelines I wrote a while ago: > https://europeanjournaloftaxonomy.eu/index.php/ejt/issue/view/1425 > In the papers, you'll find some caveats about the statistical analyses > and visualization of superimposed data and the references to the papers > covering those issues. > > I've not used CoordGen8 for ages and can't remember what it does precisely. > However, before importing the data in MorphoJ, I suggest you to rescale > the objects to their original centroid size (CS). Otherwise, that > information is lost in MorphoJ unless you import CS as a covariate. Be > careful, because MorphoJ (after redoing the superimposition) will show > you a CS variable regardless of whether you imported shape or form data. > In the first case, however, CS is just rounding error (1.00001, 0.9999 > and the like). If you import form (i.e., rescaled superimposed data), > that's not an issue and you have the correct CS. Most of the time, you > want to analyze both shape, size and maybe allometry. > Rescaling is trivial: in xls, multiply each shape coordinates for the > corresponding CS. Alternatively, if you use TPSRelw, there should be an > option for rescaling. It used to be called "restore size" (more or less) > but possibly now it's Boas coordinates. Jim will correct me. I am not > sure and cannot check now, as I have a beta version installed which is a > bit different. > > Good luck with the analysis! > Cheers > > Andrea PS I think there's a later version of MorphoJ now. > > > On 09/07/2025 20:12, 'Lydia Irons' via Morphmet wrote: > > Hi, > > > > I’d like some input on whether I’m using various geometric morphometric > > programs correctly or not. I’m a research technician working on a geometric > > morphometric project on seed beetles, and I’ve mainly taught myself how to > > use all these programs by reading various handbooks and such. Since there > > are a lot of steps, I want to make sure I’m not missing anything or doing > > something incorrectly. Here’s how I’ve been doing it, starting from placing > > landmarks all the way to running statistical analyses: > > > > 1. I use tpsUtil32 to create a tps file of all the photos I want to > > landmark using the “Build tps file from images” operation. I hit “Input” > > and choose an image file from the folder where all my images are, then I > > hit “Output” and create a name for the tps file. I then select “Setup” and > > include all the images I need, then hit “ok” > > 2. Then I upload this file to tpsDig264 and place the landmarks first, then > > the semilandmarks, for each specimen. I use the “Digitize landmarks” tool > > for landmarks, and the “Draw background curves” tool for semilandmarks. I > > make sure every specimen has the same number of landmarks and > > semilandmarks, using the “Resample curve” option to keep the number of > > semilandmarks consistent for each curve > > 3. Once I’m done landmarking, I open the tps file using Notepad and change > > the number after “ID=” to the unique specimen IDs I’ve given to each > > specimen > > 4. Then I use CoordGen8 and upload the tps file using the “load tps, > > convert curves, create SLM proto” option. I save the protocol file, then do > > a “SemiLandMark Alignment” and upload the protocol file to align all the > > landmarks and semilandmarks. Then I hit “save all” > > 5. The resulting file then has to be formatted for MorphoJ, which involves > > adding the specimen IDs to the beginning of each block of coordinates that > > corresponds to that specimen, deleting the centroid size at the end of each > > block of coordinates, making sure a paragraph mark separates each > > specimen’s block of coordinates, and adding a comma followed by a space > > between every number > > 6. Once the formatting is done, I upload this file to MorphoJ 1.07a by > > selecting “create new dataset” and selecting the data as 2 dimensions, > > selecting “yes” for object symmetry if applicable, and choosing the file > > type as “text.” Then I click “new procrustes fit” and line up the landmarks > > if there is object symmetry. I look for any outliers, generate a covariance > > matrix, extract classifiers from my specimen ID, and run whatever analyses > > I need to (PCA, CVA, ANOVA, etc.) > > > > Are there any steps I’m missing or things I’m doing incorrectly? > > > > Thank you in advance, > > Lydia Irons > > > > -- > Andrea Cardini > E-mail address: [email protected], [email protected] > WEBPAGE: https://sites.google.com/view/alcardini2/ > > -- > 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 [email protected]. > To view this discussion visit > https://groups.google.com/d/msgid/morphmet2/cb16dbac-a966-435a-a1c9-6f1ed43f8267%40gmail.com. -- 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 [email protected]. To view this discussion visit https://groups.google.com/d/msgid/morphmet2/CAAdPgfMdudnjC9uc--yWxxaAwsEzP%2Bf8SwPxhQjUayD5EgSoRw%40mail.gmail.com.
