Abstract
The purposes of this investigation were to determine the validity of various methods (available anthropometric equations and near-infrared light interactance) for estimating body fat and to develop multiple regression equations for the prediction of body fat. Thirty-eight healthy males(age: 20.87$\pm$7.17 yrs) and 12 females(19.58$\pm$2.19 yrs) underwent hydrostatic weighing to determine body fat. Anthropometric measurements were taken of height, weight, nin skinfolds and thirteen circumferences. The results obtained are summarized as follows: 1) Relative body fat determined by underwater weighing was 12.08$\pm$5.21% for the males and 17.97$\pm$5.75% for the females. 2) Circumference and skin fold that had the highest correlation with the body fat were waist girth in males and females(r=0.60, r=0.96, respectively), and subscapular in males(r=0.68) and triceps in females(r=0.96). 3) Corss-validation of 18 selected equations on males revealed total errors ranging from 3.76% to 5.06%. Among these equations, M3(Pollock et al.) demonstrated the least total error. Total error of estimation by near-infrared(NIR) was less than that of available anthropometric measurement equations. The results of the cross-validation of 12 equations on females revealed that F3(Sloan et al.) was clearly superior in accuracy of prediction. 4) Correlational analyses showed that estimation of body fat by NIR measurement seemed to be more closely associated with body fat determined by underwater weighing in women than men, in older subjects than younger ones, and in fatter subjects than leaner ones.