Percentile values are valuable statistics for representing the extreme ends of a distribution of sizes for a single human body dimension; however, their use is not suitable to every problem. When it is desirable to combine dimensions in order to construct a model of the human body or any of its parts, percentiles can create problems due to the fact that, with the exception of 50th percentiles, percentile values are not additive. This report demonstrates the seriousness of the problems associated with the use of percentiles, and describes and compares an alternative approach for representing human body size variability. This alternative, which utilizes regression equations, offers a solution which is easily accessible and demonstrable an improvement over percentiles for the purpose of creating human models.
Abstract