Maartje Boer

CHAPTER 7 198 Methods Sample Data came from the Digital Youth project: a longitudinal study among students assessing self-report internet-related behaviors and wellbeing (Van den Eijnden et al., 2018). Students were recruited through schools in urban and suburban areas in the Netherlands. Schools were selected based on the project initiator’s personal network of contacts with key persons in schools. The data include five waves of data with yearly time intervals, conducted in February-April of 2015 until 2019. In each survey round, students from previous round(s) were invited to participate, but also new students from different grades entered. For the present study, we selected four waves of data from students enrolled in 7 th grade at time of the 2015 or 2016 survey rounds, which yielded two subsets: students sampled from 2015 to 2018 ( n = 1,352) and students sampled from 2016 to 2019 ( n = 998). The two subsets were merged, such that each subset consisted of four waves that we refer to as T1 to T4. Hence, growth was modelled as a function of students’ grade, whereby all students were enrolled in 7 th grade at T1 and in 10 th grade at T4. Students who repeated a class ( n = 46) or who participated in less than two waves ( n = 885) were excluded, yielding an analysis sample of 1,419 included students. Excluded students reported higher levels of problematic SMU, lower life satisfaction, higher impulsivity, and poorer social competencies than included students, but with small effect sizes (Cohen’s D = 0.114 to 0.216). Also, the proportion of boys, adolescents attending pre-vocational education, and adolescents with an immigrant background was higher among the sample of excluded students, although these differences were very small (Cramer’s V = 0.064 to 0.109). There were few differences between the two subsets from the analysis sample: Adolescents in the second subset reported higher levels of problematic SMU, but also higher life satisfaction than adolescents in the first subset, although these differences were small (Cohen’s D = 0.171 and 0.131, respectively). Additionally, the proportion of adolescents attending pre- vocational educationwas highest in the second subset, although here too, the difference was small (Cramer’s V = 0.137). Despite these small differences, we found that the (variances of the) initial level and development of problematic