Sex estimation through the lens of cross-sectional geometric properties of the ulna

Sex estimation is an essential part of anthropological analysis in both forensic and archaeological studies, as it is vital for the construction of biological profiles from skeletal remains. The last years have seen a steady increase in the development of alternative methodologies for sex estimation, which do not depend on strictly traditional osteometric measurements. The present study focuses on the evaluation of the sex diagnostic capacity of the diaphyseal cross-sectional geometric (CSG) properties of the ulna on 215 individuals (120 males, 95 females) from a contemporary Greek population sample, utilizing support vector machine supervised learning algorithms for the classification analysis. The correlation of age-at-death on the utilized CSG variables and the effect of bilateral asymmetry in the presence of sexual dimorphism were evaluated as well. The highest cross-validated accuracy reached was 98.15%, exceeding the accuracy achieved by standard ostemeotric measurements of the ulna, highlighting the importance of evaluating and, subsequently, incorporating alternative measurements in sex estimation practices. The optimal classification model is freely available as a standalone R function, in order to facilitate the utilization of the CSG properties in forensic context.