54 Table 3.1 Sample Information per Cohort 1st cohort 2nd cohort Start of innovation School year 2017-2018 School year 2018-2019 First interview round September 2017 (n = 11) September 2018 (n = 11) Second interview round Gender July 2018 (n = same 11) 5 females, 6 males July 2019 (n = same 11) 3 females, 8 males Educational sector 4 primary, 5 secondary, 2 vocational education schools 6 primary, 1 secondary, 4 vocational education schools 3.3.3. Data collection: Interviews The first author conducted one-hour face-to-face interviews twice with every school principal, at the beginning and end of the implementation year (see Table 3.1) to study possible differences within the implementation year. We interviewed them twice to obtain a thorough understanding of the rationale of the leadership practices shown. Informed consent forms were signed before the interview started. The interview questions were developed as part of the larger research project, which aims to study context and intervention variables, including leadership practices, in relation to effects of the program. In order to explore leadership practices, in this study we used the following broad and open questions: ‘How do you see your role as a school principal, regarding the implementation of the program and in general?’, ‘Who is responsible for leadership in this school?’, and ‘What is the responsibility of teachers regarding innovation and leadership?’. We asked the school principals to describe their practices in detail and to illustrate them with examples. To decrease the researchers’ influence on the data collection (Varpio et al., 2017), interviews were audiotaped, and transcripts were written during the interview by an assistant, and member checks were conducted by asking all school principals to check their transcript. This process led to negligible changes in the transcripts of five interviews. 3.3.4. Data analysis Our analyses are inspired by the grounded theory approach (Glaser & Strauss, 1967; Thornberg et al., 2014), using constant comparative analysis by multiple rounds of coding and two cohorts of interview data. The two cohorts were treated as a split sample in the analyses (Watling & Lingard, 2012). In this way, we used the concept of saturation to reach a ‘good enough’ information power, following Varpio et al.’s (2017) criticisms on the challenges “about whether theories, data or themes can ever be truly saturated” (p.
RkJQdWJsaXNoZXIy MTk4NDMw