Maartje Boer

CHAPTER 7 206 averages across all waves instead of using only the T1 data. Also, the analysis sample yielded 1,419 adolescents instead of the preregistered 1,414, which was due to a correction on the sample selection. For the remainder, all analyses followed the preregistered procedures. The preregistration and codes for data selection and all analyses may be consulted via https://osf.io/r9t4a/. Results Descriptive Statistics Table 7.1 shows the descriptive statistics for all study variables. It shows that the observed average level of problematic SMU was low, whereas the level of SMU frequency was around the midpoint of its scale. Observed averages in life satisfaction, self-esteem, and social competencies were high, whereas attention deficit and impulsivity were low, given the ranges of the respective scales. Identifying Trajectories Average Trajectory Figure 7.2 shows the estimated average trajectory of problematic SMU and SMU f requency over time. At T1, the average reported level of problematic SMU was 1.153. The course of problematic SMU was non-linear, whereby adolescents’ level of problematic SMU first increased, but decreased after T2 ( B linear = 0.142, p = 0.015; B quadratic = -0.060, p = 0.002). Also, there was a non- linear trend of SMU f requency, whereby SMU f requency increased until T3, but decreased thereafter ( B linear = 0.366, p < 0.001; B quadratic = -0.087, p < 0.001). Model Selection Table 7.2 shows the fit indices and classification accuracy of six LCGA models. The higher the number of classes, the better the model fit in terms of the BIC and BLRT, as the BIC decreased until the final model and the BLRT p- value indicated that adding classes improved model fit compared to a model with one class less ( p < 0.001). However, for the 5- and 6-class models, the decrease in BIC was relatively small, and for the 6-class model, the LMR-LRT p- value