Sonja Mensch

99 Construct validity and responsiveness Table 1. Part of the data set after restructuring for responsiveness testing. Nx are data points Participant T0 T1 Event 1 N1 N2 Yes 1 N2 N3 Yes 1 N3 N4 No 1 N4 N5 No 1 N5 N6 No 2 N7 N8 No 2 N8 N9 No 2 N9 N10 Yes 3 N11 N12 Yes 3 N12 N13 Yes 4 N14 N15 Yes 4 N15 N16 No 4 N16 N17 Yes 4 N17 N18 Yes 4 N18 N19 No Responsiveness was analysed using two approaches: comparison of mean Movakic score-changes in intervals with and without events, and comparison of intraclass correlations of T0 andT1 scores in intervals with andwithout events. In the first approach, responsiveness would be demonstrated if the mean absolute score-change in the event group was larger than in the no event group. The difference between the means was analysed using Student’s t-test, and the magnitude of the difference was expressed as Cohen’s d effect size (small >.2, moderate >.5, large >.8) (Cohen., 1992). In the second approach, agreement between T0 and T1 scores was tested using two- way mixed model ICC’s. Responsiveness would be demonstrated if the ICC’s in the event group were lower than in the no event group, as shown by non-overlapping 95%-confidence intervals (CI). The proportion in the event group of the direction of change was calculated. Statistical analyses were performed using IBM SPSS Statistics v. 22.