Prediction accuracy of body density, lean body weight, and total body volume equations.

The purpose of this study was to determine the accuracy of percent fat estimates derived from regression equations with functions of predicting body density (BD), lean body weight (LBW), and total body volume (TBV) from anthropometric variables. BD, LBW, and TBV equations were derived from the data of 95 young, adult men (percent fat X = 13.4%). The multiple correlations for these equations were: BD, 0.86-0.83; LBW, 0.96-0.95; and TBV, 0.99. The zero order correlations between laboratory determined percent fat and percent fat derived by the BD, LBW, and TBV equations ranged from 0.80 to 0.86 with standard errors from 3.1% to 3.7%. This shows that BD, LBW, and TBV equations have similar accuracy when transformed to percent fat. The derived equations were cross validated with three additional, but diverse, samples (percent fat X = 5.1%; 16.7%; 27.1%). The cross validations results revealed that all equations exhibited similar accuracy. With samples differing in percent fat, systematic prediction errors occurred. The results further confirm population specificity of prediction equations.