Development of a hierarchical estimation method for anthropometric variables

Abstract

Most regression models of anthropometric variables use stature and/or weight as regressors; however, these 'flat' regression models can produce large errors in estimation for anthropometric variables having low correlations with the regressors. A novel method was proposed which estimates anthropometric variables in a hierarchical manner based on the geometric and statistical relationships between the variables. This hierarchical estimation method first constructs estimation structures by analyzing the dimensional characteristics and geometric relationships of the anthropometric variables and then develops regression models based on the estimation structures. The hierarchical estimation method was applied to 60 anthropometric variables (selected for the design of an occupant package layout in a passenger car) by using the 1988 US Army anthropometric survey data. The hierarchical regression models showed a 55% increase in adjusted R-2 and a 31% decrease in SE on average when compared with corresponding flat regression models.