Multivariate Methods in Engineering Anthropometry

Abstract

Engineers and scientists have known for at least 45 years that univariate descriptive statistics, particularly percentiles, are inappropriate for use in designs that involve more than one dimension. Yet, the non-existent 5th percentile woman, and 95th percentile man are commonly referred to in the literature today. Arguments against using percentiles are numerous and compelling. It has been demonstrated that: 1) they will not ensure the accommodation of the desired population, 2) they are not additive so they will result in strangely proportioned models, and 3) they result in designs that waste materials. The source of the problem is the fact that they are univariate (one-dimensional) statistics applied to multivariate (many dimensional) situations. The primary argument for using percentiles has been that there were no alternatives. In this symposium we seek to remedy that. The speakers in this symposium will describe, compare, contrast and demonstrate some of the multivariate alternatives they have been developing and using. It is hoped that this will lead the way for other engineers and scientists in the field.