Research design and analysis of data procedures for predicting body density.

Statistical methods used to develop body composition prediction equations were examined. Equations published in the 1960s and early 1970s have been called "population-specific" equations because they were developed on homogeneous samples. A major limitation of population-specific equations was a low ratio of subjects per variable, which reduces validity. The more recent method has been to develop "generalized" equations with large samples varying greatly in age and body fatness. The statistical models used for generalized equations were nonlinear regression analysis to account for the quadratic relation between body density and skinfold fat. Age was used as an independent variable to adjust for aging. Generalized equations exhibit accuracy consistent with population-specific equations and increase the generalizability of prediction. Future body composition research should focus on the cross-validation of published equations to define the limits of equation accuracy.