Reliability of multilevel functional data in resistance training biomechanics - a pointwise mixed-modelling approach

Despite the increasing use of waveform analysis methods in resistance training biomechanics research, test-retest reliability assessments remain largely focused on discrete outcome measures. Thus, this study aimed to investigate test-retest reliability of joint kinetic waveforms of a heavy lower-body strength training task by adopting a pointwise linear mixed-effects model. In total, 21 participants performed 2 sets of 5 repetitions of rear-foot-elevated split squat (RFESS) at 80% 1-repetition maximum (RM) on two separate sessions, spaced apart by 7.4 ± 1.1 days. Reliability was assessed for joint moment and vertical ground reaction force (vGRF) variables using intraclass correlation coefficient (ICC_J) and smallest detectable change (SDC_J) for J averaged repetitions. Random intercept linear mixed-effects models were applied pointwise to each curve point to acquire variance components for ICC_J and SDC_J. Joint confidence intervals and bands were constructed using bootstrap. The most reliable portions of RFESS were the middle 25-75% region (ICC_J 0.75 to above 0.90), with both the onset and end of repetition showing wide confidence intervals spanning poor-to-excellent ICC_J values. The mixed-effects models revealed that there was substantially more variance attributed to within-session than between-sessions components. Further analysis revealed that ICC_J and SDC_J values could be improved by increasing number of repetitions collected, with diminishing returns past approximately 10 repetitions. In conclusion, parts of RFESS repetition with high force demands show good-to-excellent kinetic waveform reliability in healthy young adults, whereas the start and end of the repetition suggest high uncertainty in reliability and relatively large smallest detectable change thresholds.