Maartje Boer

CHAPTER 3 72 observed in Greece and Hungary ( Δ CFI = -0.006), and the strongest increase in RMSEA was observed in Greece ( Δ RMSEA = 0.003). Age invariance was not established in Malta ( Δ CFI = -0.013, Δ RMSEA = 0.008). In the other 43 countries, age invariance was established, whereby the highest decrease in CFI and increase in RMSEA was observed in Romania ( Δ CFI = -0.007, Δ RMSEA = 0.004). Socioeconomic invariance was established in all countries because CFI decreased with not more than 0.002 (Sweden) and RMSEA decreased in all countries with at least 0.001 (Kazakhstan). The invariance analyses by country are presented in the Appendix (Tables A3.7-A3.9). Criterion Validity Table 3.5 shows the means in life satisfaction, psychosomatic complaints, and intensity of online communication via social media, by problematic SMU, as well as the effect sizes of the mean differences. Although the outcome measures show skew distributions, it is unlikely that this significantly affects the results, because large samples were used (Schmidt & Finan, 2018). Furthermore, mean differences were estimated using regression with MLR- estimation, which provides estimates robust to non-normality (Kline, 2011). In the pooled sample, problematic users reported lower levels of life satisfaction, higher levels of psychosomatic complaints, and higher online communication intensity than non-problematic users. The difference in life satisfaction and intensity of online communication between problematic and non-problematic users was small to moderate, whereas the difference in psychosomatic complaints was moderate to large (Table 3.5). The analyses by country showed that there was a negative association between problematic SMU and life satisfaction in 40 countries, with effect sizes ranging from small (Albania: β = -0.193, p = 0.021) to moderate/large (England: β = -0.682, p < 0.001). In four countries, there were no significant differences in life satisfaction (Azerbaijan, Georgia, Kazakhstan, and Republic of Moldova). The positive association between problematic SMU and psychosomatic complaints was observed in all countries, with effect sizes ranging from small/moderate (Norway: β = 0.309, p < 0.001) to large (Azerbaijan: β = 0.924, p < 0.001). The positive association between problematic SMU and the intensity of online communicationwas observed in 41 countries and ranged fromsmall (Armenia: β = 0.163, p = 0.023) to moderate/large (Switzerland: β = 0.635, p < 0.001). In two