Estimation of sex in Brazilian samples with cross-validation in populations of different regions

Abstract

Sex estimation might be influenced by ancestry and Brazilian population is cross cultural. Skulls are often used for estimating sex dimorphism. The purpose of the present study was estimate sex based on five craniometric measurements from contemporary and identified samples from two different Brazilian regions with cross-validation. Ethnic group was estimated by Nasal Index and Prosopometric Facial Index. A population sample from South region of Brazil was used to estimate a model (train) and then its performance was tested in a population sample from the Southeast region (test). In a second moment, we performed analysis inverting the order or the populations. The two samples (Southeastern-Southern) were united in one group, in order to verify the influence of this union in the prediction of sex. Models with one, two, three independent variables were created using binomial logistic regression. The results revealed that sex estimation is best performed without the interference of the variable ancestry. Three craniometric measurements allowed a more predictive model for populations with high degree of cross cultural, as the Brazilian one. United samples, Southeastern-Southern, resulted in a good prediction (AUC = 0.85), but an intermediate value between southern (AUC = 0.89) and southeast (AUC = 0.81). It is concluded that sex estimation is best performed without the interference of the variable ancestry. Three craniometric measurements allowed a good predictive model for the Brazilian Southern and Southeastern populations. Keywords: Forensic anthropology population data, Sexing estimation, Cross cultural, Craniometric measurements