Assessment of sexual dimorphism in the humerus among a Greek Cypriot population using binary logistic regression and linear discriminant analysis.

Abstract

Purpose: Determining the sex of unknown human remains is pertinent to the reconstruction of biological profiles in forensic anthropology. The Greek Cypriot population is underrepresented in forensic anthropology literature, with only a handful of sex estimation studies having been produced thus far. The aim of this research is to provide accurate and reliable methods for estimating the sex of Greek Cypriot remains to forensically evaluate unknown human remains.

Methods: This study created classification models using two statistical methods, binary logistic regression (BLR) and linear discriminant function analysis (LDA), to determine which method provided more accurate sex classification based on measurements of the humerus in a Greek Cypriot population. Additionally, cut points were calculated for use in classification. The sample consisted of 119 Greek Cypriots from the Cyprus Research Reference Collection (CRRC; 1975-2015). Four classification models were built, implementing BLR and LDA for both left- and right-side measurements. These models were analyzed using accuracy rates, receiver operating characteristic (ROC) curves, area under the curve (AUC), and Cohen's kappa.

Results: The findings revealed that all four models demonstrated good to excellent classification rates based on AUC (0.88-0.91) and accuracy rates (85.56-87.92%). Maximized summed sensitivity and specificity ratios, ranging between 1.55 and 1.76, were used to determine the optimal cut points by measurement.

Conclusion: Based on these results, BLR is a better choice to evaluate sexual dimorphism of the humerus in Greek Cypriots. Further, cut points based on individual measurements can serve as useful markers for classifying humeri by sex.