A Method for Defining and Selecting Multivariate “Percentile” Exemplars: H-Contours

Abstract

Designing to fit the “95th percentile man” can proceed once the “95th percentile man” is given some conceptual and statistical meaning. Because of the multivariate nature of the problem, there are a number of ways to define and construct such a “95th percentile man”. At present, a percentile figure is generally one which has the given percentile for height but has the rest of the body segments proportioned according to some formulae relating the segments to the height. It is well known that the such resultants are not satisfactory representatives of the population. As a result, one often tries to piece manikins together by attaching (e.g.) “5th percentile” arms onto a “95th percentile” tall figure, or otherwise combining different percentile segments in order to try to represent the considerable percentile variation in body segments existing within individuals in the population of a “95th percentile (height) man”. Methods for constructing more representative exemplars (or manikins) of a population are needed for multivariate design and analysis. This paper proposes a method based on “h-contours” which has the advantages of conceptually extending the univariate percentile concept as well as being suitable for design by constructing realistic multivariate exemplars. A multivariate Gaussian distribution is used for computational purposes. The proposed method is briefly contrasted with the “ellipse” method as proposed by others. A bivariate example is used to illustrate the concepts and computations.