Untitled

Abstract

Body composition (BC) assessment is often conducted using one of several different field techniques, which individually are considered valid tests. Anecdotal evidence has suggested, however, that some individuals may rank relatively high when assessed by one method and relatively low when assessed by another method. This inconsistency would indicate that BC assessments have poor convergent validity. The purpose of this study was to examine the convergent validity of common BC assessments using a norm-referenced approach. A total of 67 college students participated in this measurement study and had their BC assessed by each of three different tests: percent body fat (PBF) by skinfold technique (PBFSF), waist circumference (WC), and body mass index (BMI). Two different statistical procedures were used to evaluate convergent validity of the three BC assessments. First, Cohen's weighted kappas were calculated using quartiles of each BC measure. This analysis utilized three different 4 x 4 tables from all BC measure pairs. Second, Bland and Altman limits of agreement (LOA) plots were constructed on all pairs after T-score transformation of each measure. Mean (SD) values of PBFSF (%), WC (cm), and BMI (kg/m2) were 12.3 (5.0), 87.0 (8.3), 26.8 (3.5) and 23.3 (7.0), 77.1 (8.8), 24.8 (3.2) for males and females, respectively. Simple kappas showed poor agreement across the three pairs of BC assessments and ranged from .14 to .17. The weighted kappas improved to fair agreement and ranged from .32 to .38. None of the three LOA plots showed systematic bias toward a method. However, 95% LOA were wide for PBFSF vs. WC (± 28.9), BMI vs. PBFSF (± 25.9), and BMI vs. WC (± 12.3). Results of this measurement study indicate that common BC assessments have poor convergent validity among college students. These results further indicate that BC may be a multidimensional trait, requiring a specific test depending on the specific trait of interest.