Biological sex estimation with femoral dimensions: a study of an adult sample from the osteological collection of Granada (Spain).

Estimating the biological sex is one of the key objectives in a forensic investigation involving unidentified human skeletal remains. The femur is sexually dimorphic and has been extensively employed for the assessment of sex in different populations. The purposes of this article include a detailed macroanatomical characterization of the femur in a modern reference sample of the Spanish population (N = 226; Cemetery of San José in Granada, housed in the Physical Anthropology Laboratory of the University of Granada), based on 15 measurements of the proximal, distal and diaphyseal regions, and the development of univariable and multivariable models that enable the prediction of sex from complete or fragmentary femora. Univariable models were developed with C4.5, a decision-tree algorithm. Furthermore, univariable and multivariable models were generated through logistic regression (LR) with stepwise selection procedures. The models were able to correctly estimate sex in up to 91.2% of the individuals, providing high classification accuracy and low bias under cross-validation. Logistic regression models provide a probabilistic estimate of sex, that acknowledges both the Daubert standard and the fact that biological sex is not strictly dichotomous. [ABSTRACT FROM AUTHOR]