*2.3. Measures*

#### Demographic Variables Included Sex and Age

Life Satisfaction (LS) was measured using the 5-item Satisfaction with Life Scale (SWLS) [7,27]. The SWLS assesses the cognitive dimension of subjective wellbeing rated on a seven-point Likert scale, ranging from (1) strongly disagree to (7) strongly agree. A higher total score indicates higher LS (min 5, max 35). The SWLS has been used extensively and found to be appropriate for assessing LS in both adults and adolescents [7]. The Cronbach's alpha value for the present study was 0.87.

Stress was assessed using the Norwegian 30-item version of the Adolescent Stress Questionnaire (ASQ-N) [28]. The ASQ is designed to measure normative stressors that adolescents may experience in their daily life and the extent to which the stressor experience has constituted a psychological challenge for them. Items are rated on a five-point Likert scale, ranging from (1) not at all stressful or is irrelevant to me to (5) very stressful; a higher score indicates a higher stress level. The scale consists of seven dimensions covering stress related to: school performance (e.g., item: Having to study things you do not understand), school/leisure conflict (e.g., item: Not enough time to have fun), peer pressure (e.g., item: Being hassled for not fitting in), home life (e.g., item: Abiding by petty rules at home), romantic relationships (e.g., item: Making the relationship work with your boyfriend/girlfriend), teacher/adult interactions (e.g., item: Not being listened to by teachers), and school attendance (e.g., item: Abiding by petty rules at school) [28,29]. The ASQ has been evaluated in different samples of European adolescents, indicating adequate psychometric properties [30,31]. Cronbach's alpha values for the sub-scales are presented in Table 2.

Self-rated health was assessed by one item, "How is your health at the moment?" The response options were: (1) very bad, (2) bad, (3) neither good nor bad, (4) good, and (5) very good. Measuring overall subjective health among adolescents using one item has previously been used in other studies on adolescents and found to be a valid indicator of overall health [32].

#### *2.4. Statistical Analyses*

Statistical analyses were conducted using SPSS 27.0 and Stata version 17. Descriptive statistics including means and standard deviations were calculated for the scales in the study. Multiple linear regression analysis was used to investigate associations between sex, age, time point, stressor domains, and the criterion variable LS, controlled for self-rated health. Self-rated health was included in the regression model because it is a potential confounder in association with both stress [33] and LS [34]. Differences in the levels of LS according to time point were investigated with dummy variables, where the year 2011 was used as the reference category. When looking at the stressor domains, each domain was investigated separately in association with LS in the unadjusted and adjusted multivariate regression model. Interaction effects were tested with interaction terms including sex × time and sex × each of the stress domains. Effect size for the multiple regression

model was calculated using Cohens' f2 with values of 0.02, 0.15, and 0.35 indicating small, medium, and large effect sizes, respectively. The proportions of missing values for the variables of stress, self-rated health, and LS varied in the range of 2.3–5.2%. In the construction of scale sum scores, cases with missing responses at a proportion of 20% or less were included. Model assumptions for linear regression analysis were tested, with no indications of multicollinearity. The VIF values for the independent variables ranged between 1.02 and 2.93, and the average VIF was 1.80. VIF ≥ 5 to 10 indicate multicollinearity among the variables in the regression model [35]. The Breusch–Pagan test is used to test for heteroskedasticity in a linear regression model and assumes that the residuals are normally distributed. The test indicated heteroscedasticity; however, no serious violations were found because of the large sample size. The scatter plot showed a random pattern of residuals. Multivariate linear regression analysis was conducted with a listwise deletion of cases. *p*-values ≤ 0.05 were considered statistically significant.

**Table 2.** Mean scores on life satisfaction, stress, and self-rated health across time points.

