Effects of a Data Reduction Technique on Anthropometric Accommodation
Multivariate data reduction techniques such as principal components analysis (PCA), offer the potential of simplifying the task of designing and evaluating workspaces for anthropometric accommodation of the user population. Simplification occurs by reducing the number of variables that one has to consider while retaining most, e.g. 89%, of the original dataset's variability. The error introduced by choosing to ignore some (11%) of the variability is examined in this paper. A set of eight design mannequins was generated using a data reduction method developed for MIL-STD-1776A. These mannequins, which were located on the periphery of a circle encompassing 90%, 95% and 99% of the population on two principal components, were compared with the true multivariate 90%, 95% and 99% of the population. The PCA mannequins were found to include less of the population than originally intended. The degree to which the mannequins included the true percentage of the population was found to depend mainly on the size of the initial envelope (larger envelopes were closer to the true accommodation limits). The paper also discusses some of the limitations of using limited numbers of test cases to predict population accommodation.