The estimation of biological sex through the radius and ulna in Portuguese reference skeletal samples.

Sex estimation is a critical step in the identification process of human skeletal remains, alongside age, stature and population affinities. The pelvis is the most dimorphic skeletal region but the long bones also provide good estimates of sex. This study aims to develop models for sex estimation using the ulna and radius in a pooled sample of 181 individuals (107 females, 74 males) from the Coimbra Identified Skeletal Collection and the 21st Century Identified Skeletal Collection. A metric approach was implemented for sex estimation based on these measurements, resulting in fifteen models generated through logistic regression that were subsequently tested in the Lisbon Identified Skeletal Collection (N = 107; 61 females, 55 males). The resulting models were contrived in an online responsive application ( https://vertigemdasespecies.shinyapps.io/RADIULNA/ ) designed to facilitate sex predictions. In general, the models present high accuracy and low sex-bias. Overall accuracies vary between 84.0% and 95.8% under cross-validation, and between 81.9% and 90.6% in the testing sample. Among the radius-only models, the 3R model (variables sagittal diameter of the head, head-tuberosity radial length and distal epiphysis width) presents the higher accuracy under cross-validation (94.7%), while the 4R model (same variables as 3R plus the circumference of the radial tuberosity) shows the best performance in the testing sample (90.6%). In the ulna, the model with higher cross-validated accuracy is the 4U model (variables maximum length, ulnar coronoid height and circumference at the midshaft), while the 2U model (variables maximum length and ulnar notch length) performs better in the testing set (89.8%).