National samples of 141 male and 133 female highly-trained adolescent athletes were studied to derive anthropometric-based equations predicting body density. Anthropometric measures included skinfold thicknesses at seven sites, circum- ferences at 14 sites, and diameters at nine sites. Criterion measures of body density were determined by underwater weighing with corrections for residual lung vol- ume based on the oxygen dilution method. Variable selection procedures included factor analysis followed by forward-stepping regression and polynomial analysis. For both the male and female samples, two quadratic equations utilizing either the sum of three or seven skinfold measures were derived. Within the male sample, high validity coefficients (R = 0.81 - 0.82) and low standard errors (SEE = 0.0055 - 0.0056 g-ml" !) were shown with these equations. Similar results were demon- strated with the equations for females (R = 0.82 and SEE = 0.0060 g-ml~ l). Cross- validation on independent samples of male (n = 66) and female (n = 46) adolescent athletes further confirmed these findings. In the cross-validation sample of males, predicted scores were highly correlated with actual body density (r = 0.86 - 0.87) and the total error of prediction ranged from 0.0057 to 0.0061 g-ml~ l. Among the females, these values were r = 0.82 - 0.83 and total error = 0.0058 to 0.0063 g*ml-1. These results indicate that within reasonable limits of error, the sum of three or seven skinfolds may be used to make estimates of the body density of adolescent male or female athletes
W G. Thorland, G O. Johnson, G D. Tharp, T J. Housh, and C J. Cisar. "Estimation of body density in adolescent athletes" Human Biology (1984): 439-448.