American Journal of Clinical Nutrition, Vol 58, 246-246, Copyright © 1993 by The American Society for Nutrition

Erratum

Am J Clin Nutr 1993;57:8-14. On page 12, the third paragraph should read as follows: LTA was related to body density in women, and this relationship was independent from age, mass, BMI, and skinfold measurements. Consequently, LTA can be considered an important predictor of body density, particularly in healthy older women and should be included with typically measured anthropometric variables. Although LTA only accounted for 2% of additional variance in body density and only reduced the SEE by 0.3 g/L (0.11% fat), large changes in the R² and SEE were not expected because LTA was analyzed last in the hierarchical approach. This is more apparent by noting that the correlation coefficient between LTA and body density was 0.49 at the start of the first step in the hierarchical regression procedures, 0.30 by the second step (after age, mass, and BMI were entered into the model), 0.15 by the third step (after abdominal and thigh skinfold thicknesses were added to the model), and remained 0.15 at the beginning of the fourth step because none of the circumference measurements were added during the preceding step. Thus, the relatively small changes in R² and SEE associated with LTA in the prediction of body density were due to the order of analyses and not because it was a weak correlate. The justification of LTA being analyzed last was that new potential predictor variables of body density should be tested to see if they can be added to commonly used predictors in a regression model, not vice versa.