Maartje Boer

SMU AND ADHD-SYMPTOMS 137 5 SMU intensity, attention deficits, impulsivity, and hyperactivity were computed using MLR-estimation. WLSMV-estimation was used to compute factor scores of SMU problems. Factor scores for participating as well as dropout cases were calculated based on all available data on previous wave(s). For example, for respondents that dropped out in the second wave, regression methods were used to estimate factor scores at the second wave using the respondents’ available scores at the first and third wave and the estimated model parameters (B. O. Muthén, 2004). Therefore, all 543 participants were retained in the analysis. Table 5.2 shows the descriptive statistics of the factor scores for all five measures in long format ( n = 1,629). Table 5.2 Descriptive Statistics, Factor Scores (n = 1,629) M [95% CI] SD Minimum Maximum SMU intensity 0.22 [0.16, 0.28] 1.22 -2.62 2.53 SMU problems 0.14 [0.12, 0.17] 0.49 -0.44 2.14 Attention deficits 0.12 [0.09, 0.16] 0.76 -1.37 2.95 Impulsivity 0.01 [-0.01, 0.04] 0.54 -0.87 2.52 Hyperactivity 0.02 [-0.02, 0.06] 0.81 -1.17 2.89 Notes. SMU = social media use; M = mean; CI = confidence interval; SD = standard deviation. Differences between participating and dropout participants were analyzed by predicting drop-out in T2 and T3 with the computed factor scores of previous wave(s). Multivariate logistic regression (results not shown) showed that adolescents who reported high SMU intensity in T1 were more likely to dropout in T3 (OR = 1.34, p < 0.05), although this only explained a small proportion of the variance in T3 dropout (Nagelkerke R ² = 0.010). SMU problems, attention deficits, impulsivity, and hyperactivity were not related to dropout in any of the waves. Modelling Strategy Directionality can be established by examining whether changes in ADHD- symptoms induce changes in social media behaviors, and vice versa, which refers to a dynamic process that takes place within adolescents. To study these dynamics within adolescents, between-person variance should be separated from the within-person variance, because time-invariant traits on the between-person level may confound within-person dynamics. The RI-