44 Chapter 2 Analysis For the analysis of the data, we used the original Samuelowicz and Bain framework as a starting point for context-specific adaptation. The analysis consisted of relating relevant interview fragments to the original framework, while we explicitly remained open to ways to modify the framework based on additional beliefs, belief dimensions, or belief orientations identified. First, each transcript was read and re-read to get an overall sense of the way in which teaching, learning, or knowledge was conceptualised. Then ‘areas of meaning,’ text fragments that related to participants’ educational beliefs, were identified. These text fragments were labelled according to the belief dimensions and beliefs of the original framework.14 Text fragments which did not match an existing belief dimension or belief were given a preliminary code based on the content of the interview fragment. Two team members (IvdH and MO) analysed each interview independently to enhance credibility,20 using Atlas-ti qualitative data analysis software. After the initial coding, the two team members discussed the results. Demarcation rules between the dimensions as well as between the constituent beliefs within each dimension were fine-tuned during the iterative analysis process to enable consistent coding. Parallel to this process we discussed potential new dimensions with their constituent beliefs, grouping and re-grouping the preliminary coded text fragments. Repeated re-coding occurred, and the iterative process continued until all the dimensions and their constituent beliefs stabilised (‘code’ saturation, a technique to improve the dependability of the research23). This happened after eighteen interviews. We frequently negotiated our data together with a third team member (RvdR) (investigator triangulation), looking for evidence and counter-evidence within the data, to reach consensus on all the identified text fragments and to reach agreement as to which dimension they belonged to as well as their constituent belief. We determined a preliminary belief orientation of the participant holistically, i.e. based on the whole transcript, including all the labelled text fragments. To further confirm data credibility, the five interviews which IvdH and MO considered most difficult to reach consensus on were analysed independently by RvdR, and the results were negotiated within the research team. The remaining interviews were analysed by both IvdH and MO, who reached negotiated consensus. During this procedure some minor adaptations were made to belief descriptions. Finally, we re-read all the transcripts again to further ensure data dependability, and to confirm ‘meaning’ saturation,23 that is, we checked if we had harvested all the new insights from the data. At the same time, we checked the consistency of the
RkJQdWJsaXNoZXIy MTk4NDMw