Multivariate Accommodation Models using Traditional and 3D Anthropometry

Various statistical approaches have been advocated that aim at creating statistically meaningful and representative models of human variation. While they all have in common the idea to summarize the critical space needed by the user population by a discrete number of cases, substantial differences exist as to how exactly these cases are identified. The choice of statistical procedures also impacts the number of representative cases (i.e. the efficiency of the model) as well as the actual percentage of the accommodated population (accuracy of the model). The purpose of the paper is to test strengths and fallacies of some of the more commonly found approaches using real as well as simulated data. Furthermore, an extension of multivariate accommodation models to 3D coordinate data, which can be used in CAD/CAM environments, are presented.

The results suggest that while overall accommodation percentages tend to improve when the number of variables and representative cases increases, various other factors can be identified that can significantly reduce or even invalidate the model accuracy. Consequently, simplistic approaches based on multiplying variables/cases do not necessarily guarantee pertinent models. Rather, optimization strategies must be sought to reconcile model efficiency and accuracy.