Relative body fat and anthropometric prediction of body density of female athletes.

Ninety-one percent (n = 182) of the female members of South Australian representative squads in 14 sports volunteered to act as subjects. Twenty-seven percent of them had represented Australia. The underwater weighing method together with the measurement of residual volume (RV) by helium dilution were used to determine body density (BD); the percent body fat (% BF) was then computed according to Siri. A stepwise multiple regression analysis yielded a correlation coefficient (R) of 0.863 between the criterion (BD) and the best weighted sum of predictors (anthropometric variables): BD (g X cm-3) = 1.14075-0.04959 (log10 sigma triceps, subscapular, supraspinale and calf skinfolds in mm) + 0.00044 (age in decimal years)-0.000612 (waist girth in cm) + 0.000284 (height in cm)-0.000505 (gluteal girth in cm) + 0.000331 (breast girth in cm). Only those predictors which resulted in a statistically significant increase in R (p less than or equal to 0.05) were included. The standard error of estimate of 0.00597 g X cm-3 was equivalent to 2.7% BF at the mean. This equation was shown to be largely population specific. There was a range of 7.6-35.8% of BF and the overall mean 18.5% was significantly lower (p less than 0.001) than that of 23.4% obtained on a moderately active reference sample of similar age (n = 135). If group sizes of only one or two are regarded as too small for meaningful comparison, then the lowest mean of 13.5% was achieved by the long-distance runners (n = 14). The highest averages were registered by the heavyweight rowers (24.2%; n = 7) and soccer players (22.0%; n = 11). The overall average for games players (n = 107) was 19.4%.