Document Type



This paper reviews theoretical and statistical considerations for validation of several methods currently used in the determination and analysis of skinfolds and body density and their relation to body fatness. The theoretical relation of skinfolds and density to body fatness are each examined with technical and biological sources of variation in both methods explored and quantified. An estimate of variability in body density due to variation in fat-free body composition within specific populations is .006 gm/cc causing errors in estimated percent fat of 2.7%. Body water and mineral content contribute most to the variability in the fat-free body density. Skinfolds predict body density with standard errors of estimate close to that expected (.0098 gm/cc) due to biological and technical sources. The pooled standard error of estimate from 23 samples of subjects ranging from children to middle-age adults, but excluding young male adult samples, is .0091 gm/cc; for young adult males that pooled standard error of estimate is significantly less, .0070 gm/cc (12 samples). Biological factors associated with age and sex appear to be the main factors limiting generalized regression equations between skinfolds and density. Further standardization of the skinfold method is needed to determine if there are additional biological factors limiting predictability of density from skinfold equations within populations of the same sex and age. To facilitate the development of generalized regression equations for prediction of body density from skinfolds, several general principles of cross-validation studies are listed and illustrated. The validation of curvilinear equations for prediction of density in adult subjects over a wide age range appears promising, but must wait further standardization in measurement technique and selection of skinfold sites, random sampling in various populations and the calculation of several sample statistics (R, SEE, E, mean difference, and standard deviation of predicted density values) for each cross-validation study. Also, to facilitate future cross validation of skinfold equations several equations are given which offer promise of being generalized across various populations within males and females.