Chapter 4 Methods
4.6 Analysis
4.6.1 Analysis Phase 1: familiarization, developing themes, and coding
The process of qualitative data analysis been traditionally viewed as un-transparent, leading to assumptions that the method lacks rigour (Spencer et al. 2003). However, in recent years there has been a proliferation of guides to conducting qualitative analysis (Bryman & Burges 1994; Mason 2002; Ritchie & Lewis 2003). Spencer et al. (2003) identify a threefold process of analysis, involving ‘data management’, the development of ‘descriptive accounts’, and then finally, the development of
‘explanatory accounts’. Similarly, Mason (2003, p. 148) has encouraged qualitative analysists to read their data ‘literally, interpretatively, and reflexively’. Importantly, qualitative analysis should be viewed as an iterative and indeed continuous process that may commence with the initial data collection, before progressing through the coding and organizing of data, through to writing and presenting findings (Spencer et al. 2003). This was no less so for this study. The approach taken to analysis, as outlined below, drew on elements of the ‘framework approach for applied policy research’ (Ritchie and Spencer 1994; Spencer et al. 2003).
While an early phase of analysis involved reflecting on interviews and emerging themes through field notes, analysis for this project formally commenced following the collection of data. This was largely due to time pressures and other demands which meant that it was not possible to transcribe all interviews immediately after they had taken place, rather than an explicit decision to treat analysis as a separate phase. The first phase of analysis involved familiarizing myself with the totality of the data I had collected (Spencer et al. 2003). Transcribing interviews myself, rather than contracting this out, was a distinct advantage in this regard. I made regular
notes reflecting on how participants had described their experiences on tape as I transcribed. I also noted down themes that occurred in the data on a separate
document using bullet points to develop a rough hierarchy (see Appendix VI). Initially these were largely descriptive of the generality of participant’s experiences. Topics covered experiences of particular benefits or ‘anxiety over expected reform’. Though I took the decision not to include my fieldwork diaries as formal data (Mason 2002), making reference to these as I transcribed helped to trigger recollections about the interviews and assisting my emersion in the data. Again, transcribing some time after the interviews had been conducted had the advantage of allowing me to reflect on them at some distance, and this reflection enabled me to pull out overarching themes.
Once the transcripts had been completed and verified by participants, I carefully read through them all, helping me to draw out further thoughts and themes. Again, these were largely descriptive, although emergent themes such as ‘disability hate crime’ and ‘relationships with PAs’ that would later form the basis for more explanatory accounts (Spencer et al. 2003) began to appear. Finally, a two-page narrative account was developed for each participant, giving an overview of the key issues that had been discussed in that interview. The application of themes to raw data is necessary to imposing order on the rich and varied material that is generated through qualitative researching (Spencer et al. 2003). However, a downside of this approach can be that data become fragmented and divorced from their original context (Richards & Richards 1994). These narrative accounts therefore helped me to keep each participant’s ‘story’ in mind as coding and fragmentation of the data took place.
Once fully immersed in the data I drew together my notes and emergent themes document into a more structured coding matrix (Appendix VII). Themes were grouped under similar headings and connections made between different concepts and ideas. These included the descriptive a priori themes discussed above, but also reflected ideas that had emerged through the research process (Ritchie & Spencer 1994). For example, the experience of financial constraint was an a priori theme, although the treatment of different forms of income as ‘special money’ was a pattern that emerged through familiarity with the data.
The application of codes to the data was conducted using NVivo software for qualitative data analysis. Codes were applied to all transcripts in a systematic way ensuring that all of the data was fully interrogated and explored for meaning
(Spencer et al. 2003). As with all methods of analysis, using computers to interrogate qualitative data comes with both advantages and drawbacks. Technology can restrict creativity as the researcher is constrained by the limits of the software’s functionality (Richards & Richards 1994). This can also undermine analysis if coding is started before the researcher is properly familiar with the data (ibid). On the other hand, computer software can assist researchers in getting a sense of how widely experienced a particular phenomenon is, though the depth of feeling about a
particular issue is also important to capture (Spencer et al. 2003). Crucially, the use of software significantly reduces the administrative burden of fragmenting and organizing large amounts of textual data (Richards & Richards 1994). NVivo is widely used by social science researchers and many institutions offer training in how to use it. I attended two NVivo training courses before starting work on coding. I also made use of online tutorials to familiarize myself with the software and its uses. These tutorials focused on functionality: how to input and organize data, create codes, and carry out simple queries. Courses did not cover creative ways of thinking about data and developing ideas, and this was something I felt was an important gap in the training made available to doctoral researchers.