Elien Neimeijer

36 to align with all interviewers how to present and explain information to participants unambiguously. Neither the possibility of socially-desirable answers can be excluded. Consistency in answering patterns, the fact that the questionnaire contains both posi- tively and negatively formulated items and interviewers were not in any way involved in treatment suggests that the influence of social desirability was minimised. A further limitation is that we used a single itemmeasure to assess convergent validity of the GCI. This may yield biased results, because the statements corresponding to the subscales may not capture all relevant aspects of the different factors of group climate. Future studies should assess convergent validity of the GCI with a validated group climate instrument, such as the EssenCES (De Vries, Brazil, Van der Helm, Verkes, & Bulten, 2018). Also, future studies should examine concurrent validity and predictive validity of the GCI in populations with MID-BIF. Concurrent validity can be assessed by relating group climate to aggressive behaviour during treatment, such that a positive group cli- mate could be associated with fewer aggressive incidents (De Decker et al., 2017; Ros et al., 2013). Predictive validity can be established by examining the relationship between quality of the group climate and treatment outcomes (Bressington et al, 2011; Schubert et al., 2012; Tonkin, 2015). Future studies on the GCI should focus on the clustered nature of group climate measures Another important methodological limitation is that we used conventional single level CFA to examine the factor structure of the GCI. The ICCs found in the present study indicate that a substantial portion of variance can be attributed to the between-group level. Therefore, multilevel analysis is warranted (Hahs-Vaughn, 2016; Hox, 2002). The assumption is that the perception of group climate varies across individuals, and groups vary in average level of group climate. Also, it can be argued that the perception of the group climate is determined by characteristics of the group more strongly than charac- teristics of participants. An important advantage of MCFA is that the factor structure of a measure can be examined at both the within-group level and the between-group level (Huang, 2017). However, in the present study, the sample size was insufficient to conduct a multilevel confirmatory factor analysis (MCFA). Future studies on the GCI (and other group climate instruments, see Tonkin, 2015) should focus on the clustered nature of group climate measures (individuals are nested within groups). It is important to examine