Maartje Boer

VALIDATION OF THE SMD-SCALE 51 2 sets of criteria. These subgroups were labelled as normative, risky, and problematic users, respectively. Further, the criterion validity of the test score interpretations was good: In line with previous research, a higher level of problematic SMU was associated with a higher probability of reporting mental health problems, school problems, and sleep problems. Furthermore, problematic users reported the most mental health, school, and sleep problems, followed by risky and normative users. Girls, low- and medium- educated adolescents, 15-year-olds, and non-Western adolescents endorsed more problematic SMU criteria than boys, high-educated adolescents, 12-year- olds, and native adolescents, respectively. The finding that the dimensionality assessments identified one underlying factor and that all nine items substantially contributed to the factor implies that the scale measured one construct as intended, and that computing a sum-score from all nine items to assess problematic SMU is valid. It has been argued that some items may identify problematic (social media) behaviors more strongly than others (Kardefelt-Winther et al., 2017). Although the factor loadings of the nine items varied, the small observed differences in their strengths do not support this theory-driven argument. In addition, although the SMD-scale was developed as a unidimensional scale, arguably, a multidimensional factor structure would have been plausible. For example, one may argue that some criteria relate to a behavioral dimension of problematic SMU (e.g., conflict, problem), whereas others to a cognitive (e.g., preoccupation, tolerance). The finding that the unidimensional factor structure was most adequate implies that despite the potential conceptual overlap between particular criteria, together the nine criteria reflect one underlying dimension. However, to consolidate this suggestion, additional exploratory dimensionality tests on data from an extended version of the SMD-scale, that uses more items per criterion (Lemmens et al., 2015; Van den Eijnden et al., 2016), are warranted. The finding that the factor structure was measurement invariant suggests that the test scores can be used to reliably compare problematic SMU sum-scores across gender, educational levels, age categories, and ethnic backgrounds. This is an important finding since to our knowledge, no previous studies have investigated measurement invariance of any problematic SMU-scale across these four subpopulations using nationally