Comparison of four simple methods for estimating sexual dimorphism in fossils.

Estimating sexual dimorphism in skeletal and dental features of fossil species is difficult when the sex of individuals cannot be reliably determined. Several different methods of estimating dimorphism in this situation have been suggested: extrapolation from coefficients of variation, division of a sample about the mean or median into two subsamples which are then treated as males and females, and finite mixture analysis (specifically for estimating the maximum dimorphism that could be present in a unimodal distribution). The accuracy of none of these methods has been thoroughly investigated and compared in a controlled manner. Such analysis is necessary because the accuracy of all methods is potentially affected by fluctuations in either sample size, sex ratio, or the magnitude of intrasexual variability. Computer modeling experiments show that the mean method is the least sensitive to fluctuations in these parameters and generally provides the best estimates of dimorphism. However, no method can accurately estimate low to moderate levels of dimorphism, particularly if intrasexual variability is high and sex ratios are skewed.