Thomas C J, Kotze T J, Van der Merwe C A
Faculty of Dentistry, University of Stellenbosch, Tygerberg, Republic of South Africa.
Arch Oral Biol. 1987;32(4):315-7. doi: 10.1016/0003-9969(87)90028-8.
A classification of human ruga pattern disclosed interracial differences between six population groups. The conventional statistical procedures of linear and quadratic discriminant function and nearest neighbour-point-method delivered overall correct classification figures of 49.2, 45.3 and 29.1 per cent respectively. The jackknife-classification matrix for the kernel-function method delivered an overall percentage correctly classified of 61.1 per cent with an individual group figure range correctly classified of 37-100 per cent. Although computationally elaborate, this technique has led to new insights into the data compared with parametric methods.
一项关于人类皱纹模式的分类揭示了六个种群群体之间的种族差异。线性和二次判别函数以及最近邻点法的传统统计程序分别给出了49.2%、45.3%和29.1%的总体正确分类率。核函数法的留一法分类矩阵给出的总体正确分类百分比为61.1%，单个群体的正确分类数字范围为37%-100%。尽管该技术在计算上较为复杂，但与参数方法相比，它为数据带来了新的见解。
