Dear Dimitri Neaux , to my knowledge a GPA does not removes entirely the size component. You can test easily if after a GPA your data continues influenced by size computing a correlation test between the PC1 of your adjusted procrustes coordinates and the centroid size. If the correlation is positive and significant you can run a multivariate regression of your procrustes coordinates on the CS to completely remove the CS of your dataset. You can do that in Geomorph, Morpho, Shapes, etc but regarding the Shapes package please check the argument tangentcoords to see how you can use the shape space or alternatively the size-and-shape space. Hope this helps. Kind Regards

Dear Miguel Eduardo Delgado, thank you for your answer. I am fully aware that GPA keep some degree of size information (that can be removed by regressing shape on size). The idea here was more to take all the size information into account (and not a part of it) in order to the analyze the form in a dataset where both shape and size matters. Maybe it is more a technical question (where to include size it in a R array composed of Procrustes data) than a conceptual one. Best Regards

Dimitri Neaux

https://hub.packtpub.com/implementing-principal-component-analysis-r/

https://www.analyticsvidhya.com/blog/2016/03/pca-practical-guide-principal-component-analysis-python/

https://datascienceplus.com/principal-component-analysis-pca-in-r/

https://aaronschlegel.me/principal-component-analysis-r-example.html

I see, sorry for the obvious response. In this case include CS as a covariate in your data frame rather than in an array could help. I don’t know your specific case, but usually some GM software (e.g., MorphoJ) incorporates the CS as a covariate to compute different types of analyses. This is also an obvious response but maybe can helps.

Dear Dimitri,

I suggest you to explore the Morpho R package. Once you import an array, you can run the function procSym setting the argument sizeshape on TRUE (the log of CS is added to the coordinates after GPA). Another function you may be interested is the CAC (common allommetric component). Anyway, my suggestion is to explore the Form space (procSym function scale=FALSE, CSInit=FALSE).

Best,

AP

Salut Dimitri,

Personnellement je fais la PCA sur cbind(two.d.array(gpa$rotated),log10(gpa$size)). Comme ça tu intègres les coordonnées Procrustes + la taille centroide

Thanks your answers. It worked great with Antonio and Colline's solutions.