65 Pacing behavior development in adolescent swimmers 4 Table 1. Total number of swimmers and observations according to sex, performance level and event included in the analysis. Performance level limits Individuals Observations Male (100m freestyle) Elite best rSBT ≤ 103.7% 145 756 Sub-elite 103.7% < best rSBT ≤ 107.4% 501 2,472 High-competitive 107.4% < best rSBT ≤ 114.7% 1,193 5,348 Total 1,839 8,576 Male (200m freestyle) Elite best rSBT ≤ 104.1% 104 524 Sub-elite 104.1% < best rSBT ≤ 107.6% 314 1,548 High-competitive 107.6% < best rSBT ≤ 116.6% 650 2,825 Total 1,068 4,897 Female (100m freestyle) Elite best rSBT ≤ 105.2% 175 940 Sub-elite 105.2% < best rSBT ≤ 107.5% 265 1,289 High-competitive 107.5% < best rSBT ≤ 115.0% 1,219 5,155 Total 1,659 7,384 Female (200m freestyle) Elite best rSBT ≤ 104.2% 142 704 Sub-elite 104.2% < best rSBT ≤ 107.5% 315 1,455 High-competitive 107.5% < best rSBT ≤ 115.8% 795 3,253 Total 1,252 5,412 Statistical analysis Following the methods introduced by Menting et al. (2020), longitudinal multilevel models were created to describe pacing behavior as a function of age, race experience and performance group. Multilevel modelling allows for the creation of models in which repeated measures (level 1) are nested within individual swimmers (level 2), allowing the use of longitudinal data with varying number of measurements between swimmers as well as a variety in temporal spacing between measurements. Analyses were performed using the lmer4 package in R (R version 3.6.0) (R Core Team, 2019; Bates et al., (2015). Statistical assumptions (e.g., multicollinearity) were checked and outliers were screened and removed (100m: 915, 200m: 1,006). The RST per 50m section were included as dependent variables. In contrast to split times, all RST must add up to 100%. With respect to this constraint, one out of two (100m freestyle) and three out of four (200m freestyle) multilevel models were created. The remaining, free section (RST 50-100m in both events) was calculated from these models. Following that the sum of 50m sections must add up to 100%, the same predictor variables (fixed part) and variance structure (random part) had to be incorporated into each model equation.
RkJQdWJsaXNoZXIy MTk4NDMw