Esther Mertens

| 97 Classmate Influence in Intervention Observation assessments were only conducted in a subsample of the larger project. In 6 schools 14 7 th grade classes were randomly selected to participate in the observation task using an online random number generator (R&W condition: 7 classes; control condition: 7 classes). Using the same online generator, students were randomly matched in same-sex dyads within their class. Dyads were composed, as recommended by Huenecke and Waas (2010), of same-sex classmates as young adolescents affiliate mostly with classmates of the same sex. Observations took place before the start and immediately after the intervention (about 4 months later) for the R&W condition and, parallel, at the same time points for the control condition. Additionally, students completed questionnaires at baseline and, about 4 months later, post intervention. Students gave active informed consent for their participation. Parents gave passive informed consent for participation of their child. Participants In total, 152 students (76 dyads) participated in the observation task at baseline. At post measurement, 130 students (65 dyads) participated again in the same dyad as at baseline. Eleven dyads were missing due to absence of one student (absent on the day of measurement n = 6, changed school n = 4, refused to participate n = 1). Missing data of the final sample ( N = 130) was missing completely at random (Little’s MCAR test: χ 2 (18) = 18.94, p = .396). Students were between 11 and 14 years old ( M = 12.37, SD = .56). Of these students, 70 (53.8%) were boys, and 85 (66.9%) had a Western background. The R&W condition consisted of 62 students of whom 34 (54.8%) were boys with an average age of 12.34 ( SD = .63). Of these students, 30 (50.8%) had a Western background. The control condition consisted of 68 students of whom 36 (52.9%) were boys with an average age of 12.40 ( SD = .49). Of these students, 55 (80.9%) had a Western background. An ANOVA and two Chi-squared tests showed no differences between the R&W and control condition concerning age and gender. The conditions differed slightly concerning students’ background (χ 2 (1) = 12.88, p < .001, φ = .318); in the control condition more students had a Western background than in the R&W condition. There were no differences at baseline between the R&W and control condition concerning deviant and prosocial modeling and reinforcement, and outcome variables. Conditions Intervention. The theory of the intervention is based on the “R&W house”. This house consists of 5 levels representing modules in which R&W aims to increase students’ experienced safety, to learn students to deal appropriately with difficult situations, to teach about (non)verbal communication, help students to develop their 5