Application of the Optimized Summed Score Attributes Method for Sex Estimation

The optimized summed scored attributes (OSSA) method was first developed for cranial ancestry estimation (Hefner & Ousley 2014). Tallman and Go (2018) adapted this method for sex estimation with the five skull traits described by Buikstra and Ubelaker (1994) and Walker (2008). Using an Asian sample, Tallman and Go (2018) achieved moderate accuracy rates (83.7% calibration; 81.9% validation) but also high sex bias (29.1% calibration; 34.5% validation), possibly due to lower levels of sexual dimorphism in Asian populations. To further explore this novel approach to sex estimation, the OSSA method was applied to a U.S. Black/African ancestry and White/European ancestry calibration sample (N = 700). Accuracy rates were 77.4% in Black individuals and 77.2% in White individuals. Despite generally higher levels of sexual dimorphism in these groups, a high sex bias still occurred (15.4% Black individuals; -20.5% White individuals) using OSSA. The method was tested in a separate validation sample (N = 200) with accuracy of 78.0% in Black individuals (8.0% sex bias) and 70.0% in White individuals (-56.0% sex bias). When these same traits were tested with Walker's (2008) logistic regression and in the MorphoPASSE Program (Klales 2018) using random forest modeling, accuracy rates varied ,with OSSA (77.3% correct), performing slightly better than Walker's (2008) method (75.6% correct) but worse than MorphoPASSE (85.3% correct). The higher accuracy and lower sex bias in MorphoPASSE suggests that the Walker (2008) traits can be used to accurately estimate sex with statistical approaches more appropriate and robust than OSSA

Biology · Calibration · Estimation · Logistic regression · Regression · Sample (material · Sample size determination · Sexual dimorphism · Statistics · White (mutation · Autopsy Techniques and Outcomes · Demography · Forensic and Genetic Research · Forensic Anthropology and Bioarchaeology Studies · Mathematics