Using multilevel linear growth models to examine participant performance on different cardiorespiratory fitness assessments
Peter D Hart
Background: The ability of a cardiorespiratory fitness (CRF) assessment to differentiate individuals is a valuable characteristic. Little research exists, however, regarding the extent to which individual variation in CRF differs in the same magnitude across different CRF tests. Purpose: The purpose of this study was to use multilevel linear growth models to examine the inter-individual variation of within-individual variation of CRF field test scores. Methods: Data for this research came from N=131 college students attending a rural public university. Four (4) CRF field assessments were administered to each participant in random order with each yielding estimated maximal oxygen consumption (VO2max, ml/kg/min). Random intercept and random slope multilevel growth models were evaluated with CRF tests (level 1) nested within participants (level 2). CRF tests were coded 0 = treadmill, 1 = step, 2 = non-exercise, and 3 = beep and sex was coded 1 = male and 0 = female. Results: Results from the unconditional means model justified the multilevel analysis of the CRF data (ICC = .19, p < .001). A fixed slope for CRF test (b = -4.95, p < .001) was significantly related to CRF scores but did not show significant random variation. A final random intercept model was selected with significant level 2 predictors of age (b = -0.33, p < .001) and sex (b = 12.59, p < .001), significant level 1 predictor of CRF test (b = -3.41, p < .001), and significant cross-level interaction predictor of sex-by-test (b = -2.32, p < .001). The final model explained 42% and 70% of level 1 and level 2 variance, respectively. Conclusion: Results from this study show that CRF field assessments suffer from considerable variation. Sex-specific inter-individual variation showed consistent within-person variation, with males displaying a slightly steeper linear trajectory across CRF tests.