Maartje Boer

CHAPTER 9 296 problematic SMU is a behavioral addition. Although we found that moderate levels of SMU problems were persistent over time (Chapter 7), future research exploring the course of higher levels of SMU problems is important to the question whether problematic SMU identifies addiction to social media. Furthermore, we would like to underline that our studies focused on young adolescents. More research testing whether SMU problems are persistent and hamper daily life during other life phases are important (e.g., during late adolescence and adulthood). Such studiesmay set out whether the persistent and harmful nature of SMU problems as found in our research are limited to early adolescence or emerge over the whole life course. Methodological Implications The finding that the SMD-scale had good psychometric properties (Key finding 1; Chapters 2-3) implies that the scale is suited for research on problematic SMU. Researchers adopting the scale can use the scores on the scales in different ways. One strategy is to use adolescents’ sum-score of the nine items, which is indicative of the number of present symptoms, that is, the level of SMU problems. Given the distribution of the scale’s sum-score, (zero-inflated) Poisson regression techniques are required when studying the sum-score as an outcome (Atkins & Gallop, 2007). Another strategy is to divide adolescents’ sum-scores into categories. Based on latent class analysis (Chapter 2), we identified three subgroups of users: normative (no symptoms or one symptom), at-risk (two to five symptoms), and problematic (six to nine symptoms). This operationalization allows researchers to compare subgroups of users on, for example, their wellbeing. In addition, it may be promising to study SMU problems as a latent variable. The advantage of using a latent variable is that it considers that some items contributemore to the underlying concept than other items, although the differences in the contributions of the SMD-scale items were, in general, small (Chapters 2 and 3). Furthermore, latent variables take into account the measurement error of the items, which is often a more realistic representation of the data (Bollen, 2002). Also, the finding that rather than the intensity of SMU, SMU problems were negatively related with wellbeing (Key finding 2; Chapters 4-6, 8), together with the finding that these two SMU behaviors were correlated (Chapters 3, 4), is important for future research. More specifically, to improve