Esther Mertens

| 151 General Discussion for decreasing victimization in the class, amplified by dyadic mutuality. This means that interventions addressing classroom victimization, including R&W which was not effective in decreasing victimization, might be optimized by an increased focus on classmates’ prosocial modeling and improving dyadic mutuality between classmates. For instance, to stimulate classmates’ dyadic mutuality interventions could organize fun exercises in the class in which classmates have to work together. Even though the results of the current dissertation did not confirm the theory of R&W which suggests that classmates’ modeling and reinforcement are mediators in the intervention, the results did provide useful theoretical insights concerning a potential mechanism of change that can be used to optimize interventions (Kazdin, 2007). Studying mechanisms of change is thus crucial for both theory and intervention development, and can be examined even if a direct effect of the intervention on an outcome is absent (O’Rourke & MacKinnon, 2018). The second way in which I studied working mechanisms in interventions was through examining intervention components. The results of the meta-analysis indicated that components related to stronger intervention effects were not always commonly implemented, and conversely, some commonly implemented components were related to weaker intervention effects. Note, however, that these relations between components and intervention effects were based on correlations. Components were coded as either present or absent in the interventions included in the meta-analysis. Subsequently, the general effectiveness of interventions with this component is compared to the general effectiveness of interventions without this component. Correlational results are inherent to this type of meta-analysis and results should therefore be used for hypotheses generation rather than for drawing firm conclusions about the (in)effectiveness of individual components. Notwithstanding the correlational nature of these relations, the results suggest opportunities for optimizing interventions by adding components related to stronger effects and eliminating those related to weaker effects (Michie, Fixsen, Grimshaw, & Eccles, 2009). For instance, in line with previous research (Boustani et al., 2015; Cuijpers, 2000), my results showed that teaching students self-awareness and problem solving by means of active learning methods was related to stronger intervention effects. Hence, the results identified these components as potentially important components to include in interventions. As R&W already includes these components, it seems obvious that these components should be maintained in the program. In contrast, results frommy meta-analysis also showed that teaching students assertiveness and emotion regulation, and using group discussion were related to weaker intervention effects. These findings suggest that decreasing the use of these components might optimize interventions. R&W also includes these components, thus it seems worthwhile to consider whether the use of these components should be decreased in order to improve its effectiveness. However, optimizing interventions based on associations between components and intervention effects is less straightforward than it seems. Whether a component is related to intervention effects might depend on the circumstances of implementation 7