FORDISC 3.0 (FD3) is an analytical program distributed by the University of Tennessee that employs discriminant analysis to assist in assessing sex, ancestry, and stature from unidentified skeletal remains. While American researchers regularly use FD3, the program has seen less use elsewhere primarily due to the fact that non-American groups are not well-represented in the database. The purpose of this paper was to test the accuracy and applicability of FD3 on South African groups. The sample contained 226 individuals: 86 blacks (55 male, 31 female) and 101 whites (57 male, 44 female) from the Pretoria Bone Collection along with 15 suspected 'Griqua' (4 male, 11 female) and 24 African farm workers (18 male, 6 female) from the National Museum in Bloemfontein. Age ranged from 21 to 97 years; and specimens showing cranial pathology were excluded. Standard cranial measurements were taken, and only specimens with 15 or more measurements were used. Data were entered in FD3 and separate discriminant functions were constructed for each specimen, comparing it to 4 American subgroups (white-black, male-female). The groups that each specimen was placed in and the posterior probability (PP) of membership in that group were also recorded. Approximately 73% (164/226) of the sample was classified correctly (sex and ancestry). Black females performed the worst (60%), while white females performed best (82%). Misclassification occurred more often for sex than ancestry, eg, black males were often misclassified as black females and vice versa. Average PP for the sample was 0.87, implying that FD3 classified South Africans with a high confidence level. Errors in assignment are not mishaps of the program or statistical techniques; rather FD3 datasets are not representative of the South African population. While FD3 appears relevant for use in South Africa, the addition of South African data to the program is warranted.
Poster presented at the University of Pretoria Health Sciences Faculty Day, August 2008, Pretoria, South Africa.