Estimating ancestry in South Africa: a comparison of geometric morphometrics and traditional craniometrics

In ancestry estimation of South African individuals, non-metric morphological trait assessment has not proven useful and previous results in FORDISC 3.0 leave room for improvement. The accuracy rates of software programs FORDISC 3.1 and 3D-ID were compared for ancestry estimation based on cranial data of black and white South Africans using discriminant function analysis. Cranial landmarks were digitized using a Microscribe G2 for geometric morphometric analysis in 3D-ID, and traditional craniometric measurements for use in FORDISC were calculated using the data collection software 3Skull. Data was collected from a total of 385 individuals (186 black and 199 white crania) from the Pretoria Bone Collection. Overall accuracy rates of 75.6% using FORDISC 3.1 and 63.1% using 3D-ID were obtained for black and white South Africans. Incorrect estimates were more often due to misclassifications of sex rather than ancestry, reflecting the decreased amount of sexual dimorphism in South African populations when compared against American populations, discussed previously. Black South Africans were more often classified correctly in FORDISC 3.1, and white South Africans were more often classified correctly in 3D-ID. Low sample size in comparative databases and broad ancestral differences between South Africans and the proxy