The Discriminant Power of Specific Physical Activity and Dietary Behaviors to Distinguish between Lean, Normal and Excessive Fat Groups in Late Adolescents

Physical activity (PA) and dietary behaviors (DBs) are crucial determinants of body mass composition. This work is a continuation of the previous study of PA and DBs patterns in late adolescents. The main aim of this work was to assess the discriminant power of PA and dietary behaviors and to identify the set of variables that discriminated participants with low, normal, and excessive fat the most. The results were also canonical classification functions that can allow the classification of individuals into adequate groups. A total of 107 individuals (48.6% male) participated in examinations, which used the International Physical Activity Questionnaire (IPAQ) and Questionnaire of Eating Behaviors (QEB) to assess PA and DBs. The participants self-reported body height, body weight, and BFP, with the accuracy of the data confirmed and empirically verified. Analyses included the metabolic equivalent task (MET) minutes of PA domains and intensity, and indices of healthy and non-healthy DBs, calculated as a sum of the intake frequency of specific food items. At the beginning, Pearson’s r-coefficients and chi-squared tests were calculated to study various relationships between variables, while the main considerations were based on discriminant analyses conducted to determine the set of variables with the most power to distinguish between lean, normal, and excessive body fat groups of participants. Results showed weak relationships between PA domains and strong relationships between PA intensity, sitting time, and DBs. Vigorous and moderate PA intensity related positively to healthy behaviors (r = 0.14, r = 0.27, p < 0.05), while sitting time related negatively to unhealthy DBs (r = −0.16). Sankey diagrams illustrated that lean persons displayed healthy DBs and low sitting time, while those with excessive fat had non-healthy DBs spent more time sitting. The variables that effectively distinguished between the groups include active transport and leisure time domains alongside low-intensity PA, represented by walking intensity and healthy dietary behaviors. The first three variables participated significantly in the optimal discriminant subset (p = 0.002, p = 0.010, p = 0.01, respectively). The discriminant power of the optimal subset (contained four above-mentioned variables) was average (Wilk’s Λ = 0.755) and determined that weak relationships between PA domains and DBs resulted from heterogeneous behaviors and mixed patterns of behaviors. Identifying the trajectory of the frequency flow through specific PA and DBs allowed for well-designed tailored intervention programs to improve healthy habits in adolescents. Therefore, identifying the set of variables that discriminate the most between lean, normal, and excessive fat groups is a suitable target for intervention. The practical achievements are canonical classification functions that can be used to classify (predict) participants in groups based on the three the most discriminating PA and DB variables.