Sex estimation from modern American humeri and femora, accounting for sample variance structure

A new procedure for skeletal sex estimation based on humeral and femoral dimensions is presented, based on skeletons from the United States. The approach specifically addresses the problem that arises from a lack of variance homogeneity between the sexes, taking into account prior information about the sample's sex ratio, if known.

Three measurements useful for estimating the sex of adult skeletons, the humeral and femoral head diameters and the humeral epicondylar breadth, were collected from 258 Americans born between 1893 and 1980 who died within the past several decades.

For measurements individually and collectively, the probabilities of being one sex or the other were generated for samples with an equal distribution of males and females, taking into account the variance structure of the original measurements. The combination providing the best estimates correctly classifies 88.3% of the skeletons, with 10.8% considered unknown and 0.9% assigned to the wrong sex.

Probabilities of correct assignments are a better means of categorizing individuals as male or female than the sectioning points commonly used in skeletal studies. That is because it is possible to estimate the observer's certainty that the individual represented by measured bones was one sex or the other. A computer program is available that simultaneously considers samples of unequal sex composition. It is useful when there is contextual information available about the nature of skeletal samples (e.g., a mass burial from a battle or genocide).