morphmet wrote: > Dear morphometricians, > at present I'm working on some lepidopteran sibling species. Being very > difficult to identify these species not only by external features but > also by genitalia, I want analize (and quantify) the shape variation of > some anatomical parts. I'm asking for help about the better method to > compare individuals. I have to compare a sample of (many) undetermined > individuals to some specified specimens (typical specimens). Could I
I don't know what a "typical specimen" of anything is, but if you had a reasonable sample of known individuals, you could compare the Procrustes distance (or BE or whatever) of an unknown individual to the mean of that sample. The proportion of known individuals with equal or greater distance (or whatever) from the mean as the unknown (including that individual) would be the probability that that individual was sampled from that population (regardless of distribution). With more than one individual you might guard against Type I rejections with a Bonferroni correction, though you could quickly Bonferroni yourself out of business with a nearly complete loss of power. A sequential Bonferroni might be appropriate. -dslice > employ tpsSplin software to analyse my data? I would use my typical > specimens as reference instead of computing it. > Is it allowable to make such comparison? I would also know if there is a > value that I can employ to quantify the variation among my > reference-specimens and the set of data. In the report three values are > computed for each specimen. I'm thinking to use only d (Procrustes > Distances), or it is better use the three values together (i.e., Bending > Energy, Procrustes Distance and angle)? Thank you very much for any > suggestion. Angela Roggero > -- Dennis E. Slice Department of Anthropology University of Vienna ======================================================== -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
