Maartje Boer

THE COURSE OF PROBLEMATIC SMU 203 7 adolescent populations (E. Rose et al., 2012). We used amodifiedDutch version of the subscale (Straathof & Treffers, 1989), whereby respondents were asked, for example, whether they find it difficult to form friendships to which they can count on (1 strongly disagree to 5 strongly agree ). Scores were recoded such that high values indicate high levels of social competencies, after which mean scores were computed. Cronbach’s alpha ranged from 0.600 to 0.709. Controls The analyses controlled for several time-invariant characteristics, including gender ( boy or girl ), educational level ( pre-vocational , intermediate , or pre- university) , and immigrant background ( immigrant or non-immigrant ). Educational level was determined based on the respondents’ most recent reported level of education. Immigrant background was established based on the country of origin of the respondents’ parent(s), whereby response options were Netherlands , Suriname , Netherlands Antilles , Morocco , Turkey , and other country. These countries were selected because a large share of the immigrant population in the Netherlands come from these countries due to colonial past with and a history of labor migration to the Netherlands. Adolescents with at least one parent from a different country than the Netherlands were defined as adolescents with an immigrant background. Analytic Approach Identifying Trajectories We adopted Latent Class GrowthAnalysis (LCGA) usingMplus 8.6 (L. K. Muthén &Muthén, 2017b). LCGA explores heterogeneity of growth trajectories within a population by classifying individuals into subgroups based on their response patterns (Jung & Wickrama, 2008). It tests several class solutions, whereby each class represents a growth trajectory indicated by an intercept, slope, and quadratic term estimated from multiple repeated measures. Respectively, these three growth parameters denote the average level of problematic SMU at T1, the average change over time, and whether there is non-linear change. In LCGA-models, the variances and covariances of the growth parameters are constrained to zero, which imposes that individuals within a class have similar growth trajectories.