Maartje Boer

CHAPTER 3 66 between problematic SMU and positive mental wellbeing, such as life satisfaction, and a positive moderate association between problematic SMU and negative mental wellbeing, such as depression (Huang, 2020; Marino et al., 2018b). Reviewstudies onproblematic SMUand the frequency or time spent on SMU (including activities such as browsing, chatting) show a small tomoderate association (Frost & Rickwood, 2017; Parry et al., 2020), which may also apply to the relation between problematic SMU and online communication intensity. Hence, appropriate criterion validity would be established when problematic SMU was negatively related to life satisfaction with small to moderate effect size, positively to psychosomatic complaints with moderate effect size, and positively to the intensity of online communication with small to moderate effect size ( p < 0.05). Associations were examined using linear regression where problematic SMU predicted life satisfaction, psychosomatic complaints, and online communication, while controlling for gender, age, and socioeconomic status. Estimates of problematic SMU were standardized to interpret their effect size. As the problematic SMU scores were dichotomous, estimates were STDY-standardized (0.2 = small, 0.5 = moderate, 0.8 = large effect size) (Cohen, 1988; L. K. Muthén & Muthén, 2017a). Technical Details Missing data were imputed using Stata 13.0 (StataCorp, 2013). Analyses were conducted on the imputed datasets with Mplus 8.5 (L. K. Muthén & Muthén, 2017b). The CFAs, internal consistency, and measurement invariance analyses were conducted using Weighted Least Square Means and Variance Adjusted (WLSMV) estimation with a probit regression link, as appropriate for analyses with categorical outcomes (Rhemtulla et al., 2012). Regression analyses from the criterion validity analysis were conducted with Maximum Likelihood with Robust standard errors (MLR). In all analyses, standard errors were corrected for clustering of adolescents within schools or classes. For some countries, the analyses were conducted using sample weights to adjust for sample distribution differences with the respective population. Analyses by country were conducted with the MplusAutomation- package in RStudio 1.2.5042 (Hallquist & Wiley, 2018; RStudio Team, 2021). All codes related to the analyses may be consulted via https://osf.io/bgkec/. The analyses were not pre- registered and therefore, results should be considered exploratory.