Maartje Boer

THE COURSE OF PROBLEMATIC SMU 211 7 To gain amore detailed understanding of the classes, we report on some of the class comparisons using the z- scores of the parameter differences (Table 7.3). The intercept of problematic SMU differed significantly between all classes, except between Classes 1 and 2. This means that except for these two classes, all classes had different levels of problematic SMU at T1. The intercept of SMU frequency differed across all classes, except for Classes 2 and 4. Also, the non-linear trends of SMU frequency in Classes 1 and 4 were not significantly different. Figure 7.3 Average Co-Trajectory of Problematic SMU and SMU Frequency by Latent Class, n = 1,419 Note. Vertical bars denote the 95% confidence intervals. Predictors of Trajectories Table 7.4 displays the differences in adolescents’ demographic characteristics, subjective wellbeing, self-control, and social competencies by class. We examined whether these factors predicted class membership using multivariate multinomial regression. Given that Class 4 had the most members, we conducted the multinomial analysis using Class 4 as the reference group. Hence, estimates from this analysis indicate the extent to which, for example, higher levels of attention deficits, increase the probability