Sex estimation of femur using simulated metapopulation database: A preliminary investigation

Accurate and reliable methods of sex estimation are essential in forensic practice. In the literature, both the population based and non-population specific standards are existing. In the present work, the summary statistics of 3 femoral standard measurements (joints and length diameters) were collected from 13 published studies of modern populations. The metric data of each population were regenerated using truncated normal distribution approach. A simulated metapopulation database was constructed using the combined metric data of 2275 femur bones. Despite interpopulation differences, the calculation of stable non-population specific discriminant functions (DF) was possible using the most discriminatory variables selected by stepwise analysis. The functions were tested on raw data of 6 populations which were obtained from authentic sources. The pooled DFs produced lower but comparable accuracies to the population specific functions using fewer number of variables. The total allocation accuracies of the temporally distant Terry collection did reach better than 90% in both American whites and blacks while in the contemporary Turkish and Hispanic samples were more than 85%. In Asian populations, the total accuracies ranged from 69.7% to 84.9%. Nevertheless, raising the posterior probability threshold to 0.95 increased the total allocation accuracies above 95% in all populations using the pooled populations function of femur joints diameters. This function permitted the classification of one third to more than half of the total sample size in each population which suggest a considerable practical value in estimating sex of unknown individuals without a priori recognition of population affinity.