Data analysis

Thematic analysis or thematic coding approach (hereafter thematic analysis) was adopted in this case study to analyse the interview data. According to Braun and Clarke (2006), Fereday and Muir-Cochrane (2008), Guest et al. (2012), and Vaismoradi et al. (2013), thematic analysis leads to the identification of patterns from the data and helps to group related emerging categories together to develop overarching themes. Moreover, unlike typical

76

grounded theory approach, which identifies themes from within the data only by following a strict set of rules (Glaser and Strauss, 2009), and phenomenology which gives voice to the participants with no input from the researcher in data generation and analysis (Guest et al., 2012), thematic analysis identifies patterns by using a holistic approach in an iterative/non-linear way that allows ‘to be open to whatever would emerge from the data’ (Light and Dana, 2013). Moreover, thematic analysis is not tied to any particular theory or epistemology, thus it offers flexibility and versatility in dealing with complex research situations (Braun and Clarke, 2006 ; Vaismoradi et al., 2013 ; Aronson, 1994 ; Bourgeois and Rosenthal, 1983 ; Joffe, 2011).

While thematic analysis is more flexible than many other methods, it requires some rules to be followed. Braun and Clarke (2006) and Saldaña (2009) have suggested six steps to be followed to ensure the credibility of the coding/analysis process. These, along with the equivalent steps taken in this research, are given in the following table.

Table 9: The six step process in thematic analysis/coding

Phase Description of the process Equivalent steps taken in this research

1.

Familiarisation with the data

Transcribing data (if necessary), reading and rereading the data, and noting down initial ideas.

Interviews were transcribed using Dragon. Each interview was proofread, and the data were organised in NVivo. This was followed by first-phase of coding.

2. Generation of initial coding

Coding interesting features of the data in a systematic fashion across the entire data set, and collating data relevant to each code.

Initial coding began after half a dozen of the interviews were conducted. All of the interviews were subsequently coded.

3. Searching for themes

Collating codes into potential

themes, and gathering all data relevant to each potential theme.

Nodes were created and collated under tree structures during both first and second phases of coding.

4. Reviewing themes

Checking if the themes work in relation to the coded extracts and the entire data set, and generating a thematic map of the analysis.

Revisited the second phase of coding, analytic memos were written and attached to the corresponding nodes and child nodes that helped in generating the thematic map in coding phase three.

5. Defining

and naming

themes

Ongoing analysis to refine the specifics of each theme, and the overall story the analysis tells, as well as generating clear definitions and names for each theme.

Nodes were collated, merged, and in some cases, deleted. Memos were also revisited, and realigned. The thematic map was updated leading to the identification of broader themes, which were labeled with representative names during the third phase of coding.

6. Writing up

Selection of vivid, compelling

extract examples, final analysis of selected extracts, relating back the analysis to the research question, and literature, producing a scholarly report of the analysis.

Vivid and compelling extract examples were selected for the themes, and the findings chapters were written based on the exemplary quotes that were taken from the data.

77

In addition, I am aware of the potential disadvantages of adopting thematic analysis. According to Roulston (2001), coding and analysing textual interview data using thematic analysis without taking into account the theoretical notion of reflexivity will produce only a naive account of the data. Braun and Clarke (2003) identified four possible weaknesses in thematic analysis: (1) the risk of describing data extracts without going beyond their specific content and without analytic narrative of the data; (2) using questions from the interview guide as the only source of themes without any analytic work done in identifying themes across the entire dataset; (3) lack of clarity of the relationship between the themes, i.e., the presence of too much overlap and/or the lack of internal coherence and consistency between the themes; and (4) a mismatch of the actual data that is collected and the claims made in the interpretation of the emergent themes, primarily due to the absence of compelling examples from the data to support the claims. The above potential weaknesses of thematic coding have been minimized in this study by adopting the following four strategies.

First, data extracts have been critically analysed in a way that makes sense in light of the participants’ voices. Second, the questions in the interview guide were not the only questions asked during interviews, several probing and prompting questions were also asked depending on the contexts and the flow of the interviews. Moreover, as an inductive approach was employed in data analysis, the themes that were identified are grounded in the data, not in the questions of the interview guide. Third, while connections were found between and across themes (will be discussed later in Chapter 10), the themes are distinctively different from each other, there is little overlap between them. Fourth and final, the presentation of the findings (cf. Chapters 5 to 9) is informed by compelling quotes from the interview data. Thus, there is no mismatch between the data that were collected and the claims that were made.

That being said, other qualitative data coding and analysis methods (such as discourse analysis, phenomenological analysis, and grounded theory) may be exposed to a variety of potential weaknesses (including the above), if the researcher is not actively involved in data collection, if data is not analysed

78

thoroughly and analytically by asking critical questions about the data (Charmaz, 2014 ; Bazeley, 2009), if the physical, sociological, and psychological situations in which data were generated are not considered (Silverman, 1993), and if appropriate attention is not given to what participants said and the way they said what they said (Bazeley and Jackson, 2013). In short, like any other data analysis technique, thematic analysis inherits some weaknesses.

Thus, adopting thematic analysis is believed to be an appropriate data analysis method in the current case study, because it offers flexibility and is found to be more aligned with the overall research approach that is adopted in this study than other coding and analysis methods. Specifically, the way inductive thematic coding is carried out in the current study (discussed below) matches with the notion that participants’ viewpoints reflect the reality that can be critically analysed to make sense of their unique situations.