A robust comparison of biological shapes.

摘要: SUMMARY Localized differences in the form of two related animal skeletons are more effectively determined when resistant fitting techniques used rather than least squares. The repeated median algorithm is introduced. methods tested by comparing hominid skulls with those other primates and male ostracode female. In each case fit superior to squares fit, that difIerences readily identified agl-ee closely structural perceived on biological grounds. le llltroductioll Studies allometric growth morphological change often run into difficulties identifying measuring location extent relative shape deformations. A reference system needed for these deformations and, absence an inherent fixed system, one must be estimated. problem estimating morphology provides elegant setting comparison recently developed robust statistical techniques. Being robust, new protected against departures from assumptions analysis, such as independent, identically normally-distributed errors; being they also potentially strong influences atypical or incorrect data values. Two shapes can rarely superimposed perfectly; different criteria will generally yield results. By allowing regions large have a impact minimize true thereby obscure them. technique, however, limits influence resulting close similar not relatively deformed regions. this way, help identify similarities diflerences methods. When Sneath (1967) basis some first quantitative estimates deformation so extended fundamental descriptive work Thompson (1917), useful were available. did, fact, consider using available methods, but concluded 'Alternative less affected 'wild' values attractive practice'. Since then, development theory reported Huber (1972, 1981), Hampel