Literature Review
Chapter 3 Methodology
3.5 Research design
3.5.7 Analysis of interview data
Data analysis in a phenomenographic study is an ‘iterative’ and ‘interpretive process’ which requires careful reading and re-reading of the interview transcripts in order to identify and group essential aspects of the participants’ experiences into categories of description. This “allows each utterance to be understood from the participant’s perspective” (Cope, 2004, p. 6). Notably, there are divergent views among phenomenographers concerning who conducts the analysis (an individual or a team); managing a large data pool during analysis (whether selected quotes, large sections, or the whole transcripts should be focused on); categories of description; outcome space;
and validity and reliability checks (Akerlind, 2012, p. 120-121; Richardson, 1999). With regards to categories of description, the arguments centred on whether categories should emerge from the data (discovery) or be constructed by the researcher. The idea of pre-construction of categories as against the emergence of categories has been criticised (Richardsson, 1999). In support of this argument, Akerlind (2005, p. 323) states that:
Phenomenographic interviews are typically audio taped and transcribed verbatim, making the transcripts the focus of the analysis. The set of categories or meanings that result from the analysis are not determined in advance, but ‘emerge’ from the data, in relationship with the researcher.
On the contrary, other phenomenographers, such as Sandberg (1997), support the view that the researcher should construct the categories of description in relation to the data. According to McKenzie (2003), the construction of categories should be a:
…reflexive process whereby the researcher constantly checks any potential interpretations against the data itself, and maintains a critical awareness of their prior knowledge at all stages in the research process.[…] The researcher is constantly reflecting on whether
77 interpretations relate to the experiences of the interviewees and not simply to the researcher’s prior experience (p. 92).
In line with this view, Mann, Dall' Alba & Radcliffe (2007) suggest that data analysis in phenomenography be both a ‘discovery’ and a ‘construction’ process.
In my thesis, I have chosen to follow Stoodley (2009) in the use of “selected quotes, as an individual researcher… [and for my study to be] validated by communicative checks” (p. 71). I also align with the Mann, Dall' Alba & Radcliffe (2007) view about data analysis and categories of description. Thus, the categories of description arrived at in my thesis are not pre-constructed.
On the contrary, they are arrived at through a ‘discovery’ and ‘construction’ process; and are based on the following phenomenographic steps identified by Dahlgren & Fallsberg (1991, p. 152);
Sjöström & Dahlgren (2002, p. 341) and described by Khan, (2014, p. 38-39):
(a) Familiarisation step: in this step, the researcher reads through the entire transcripts several times to become familiar with the contents and to also correct errors in the transcripts.
(b) Compilation step: the second step involves compilation of the answers from all the participants to a certain question and taking note of similarities and differences. The main task here is to identify the most significant elements in the answers given by each informant.
(c) Condensation step: the third step requires the selection of meaningful quotes/excerpts that carry the main idea in the answers provided in the transcript and leave out parts of the answers that are not necessary.
(d) Preliminary grouping: is the fourth step where the researcher identifies and classifies similar answers into preliminary groups.
(e) Preliminary comparison of categories: this fifth step requires the researcher to revisit and revise/regroup the initial categories and try to compare and differentiate between one category with another.
(f) Naming the categories: The sixth step consists of naming the categories to emphasize their essence.
(g) The outcome space: this is the last step of the data analysis where the researcher discovers/constructs the outcome space which describes the internal relationships and the qualitatively finite number of ways in which a given phenomenon has been experienced.
In most cases, the outcome space is presented in hierarchical order.
To achieve this, the audio-taped interviews were listened to several times and then transcribed verbatim (Stamouli & Huggard, 2007, p. 184). Thereafter, the iterative process began, whereby, I
78 read through the entire transcripts several times and made some corrections (mostly of any typographic errors). After that, I employed the Nvivo 10 – a software developed to facilitate the organisation and management of a large amount of qualitative data (Richards, 2005). Some of the critical features of the Nvivo software is that it allows for many types of qualitative data to be imported into the software; and for the data to be coded and categorised into different nodes and sub-nodes. In Nvivo, the node serves as a virtual container for storing coded texts (O‘Neill, 2013).
Hence, I imported all of my interview transcripts into the software to begin the compilation process as shown in Figure 2 below, with one of the interview transcript displayed on the right for coding purposes.
Figure 2. NVivo 10 - a software structure showing imported interview transcripts
To compile the interview transcripts, I coded all responses to certain questions in the form of nodes. For instance, I created a node for all of the responses to the questions relating to the participants’ supervision experiences at the initial stage of the research process, which I labelled as ‘experiences of the initial stage of the research process’. The next step of the data analysis was the condensation step. In this step, I coded and selected relevant and meaningful parts of the texts of the initial nodes that contained the main ideas in terms of the participants’ responses/answers
79 which relate to a specific theme. For example, within the node labelled ‘experiences of the initial stage of the research process’, I created themes (sub-nodes) like, “participants’ conceptions about topic selection process” and a sub-sub-node labelled “frustration in working out research topics in supervisors’ domain” which were all supported with relevant coded texts/excerpts. This was meant to ensure that all the themes/groupings in a given node were “furnished with illustrative quotes”
(Smith, 2010, p. 125). Figure 3. displayed the theme (node) “participants’ conceptions about topic selection process” on the left-hand side and the excerpts selected/extracted to support the theme on the right-hand side .
Figure 3. NVivo 10 - a software structure showing nodes/sub-nodes on the left and supportive quotes on the right
The themes were then classified into preliminary groups, based on their similarities and the context of the original transcripts the excerpts were drawn from. After this initial grouping, my attention shifted from the individual to the collective meaning expressed by the entire group. Thus, as argued by Marton, “each quote has two contexts in relation to which it has been interpreted; first the interview from which it was taken and second the “pool of meanings” to which it belongs” (Marton 1986, p. 43). The next step was the preliminary comparison of the categories, where I looked out for similarities and differences in the preliminary groups. It was at this point that I began to
80 consider how the final categories could be presented in an orderly and hierarchical manner (in terms of constructing the categories). This took the iterative process of reading through the transcripts again, grouping and regrouping the preliminary categories. At the end, three stable categories of description were settled-on and named category one – supervision as apprenticeship/power-like relationship; category two – supervision as transacting the social; and category three – students’ yearning for a positive relationship. Although the naming of the categories was my abstraction as a researcher, the categories described as closely as possible the understanding/experience of the research supervision phenomenon from the perspectives of the participants. These three categories of description form the outcome space, which describes the qualitatively limited number of ways in which the participants experienced research supervision.
Although three categories in the outcome space were logically related, the hierarchical levels and inclusive structure of the categories was only present in category one and two, as category three did not fit in as the third hierarchical level.