Comparing two methods of univariate discriminant analysis for sex discrimination in an Iberian population

Abstract

This study assesses the performance of two analytical approaches to sex discrimination based on single linear variables: discriminant analysis and the Lubischew's test. Ninety individuals from an archaeological population (La Torrecilla-Arenas del Rey, Granada, southern Spain) and 17 craniometrical variables were included in the analyses. Most craniometrical variables were higher for men. The bizygomatic breadth enabled the highest level of discrimination: 87.5% and 88.5%, using discriminant analysis and Lubischew's test, respectively. Bizygomatic breadth proved highly dimorphic in comparison to other populations reported in the literature. Lubischew's test raised the discrimination percentage in specific craniometrical variables, while others showed a superior performance by means of the discriminant analysis. The inconsistent results across statistical methods resulted from the specific formulation of each procedure. Discriminant analysis accounts both for within-group and between-group variance, while Lubischew's test emphasizes between-group variation only. Therefore, both techniques are recommended, as they provide different means of achieving optimal discrimination percentages.