CHAPTER THREE METHOD
3.2. Semi-structured Interviews
3.2.7. Data analysis process
The researcher followed the guidelines provided by Braun and Clarke (2006) to analyse the data in this part of the study (Braun and Clarke, 2006). Braun and Clarke (2006) offer an outline guide with six phases of analysis: familiarising the researcher with the data; generating initial codes;
searching for themes; reviewing themes; defining themes and naming themes.
3.2.7.1. Phase one: familiarising the researcher with the data
The first phase of the data analysis was to familiarise the researcher with data. According to Braun and Clarke (2006), it is important that the researcher familiarises him/herself with the data to the extent that he/she is familiar with the depth and breadth of the content (Braun and Clarke, 2006). The transcription of verbal data is considered to be one of the methods which help the researcher to develop a far more thorough understanding of the data (Braun and Clarke, 2006).
Some researchers (Bird, 2005) consider the transcription process to be a key phase of the data analysis within interpretive qualitative methodology. In this research, all audio-recordings of all interviews were transcribed by the main researcher. The literature reports that as there is no single way to conduct a thematic analysis, there is no set of guidelines to follow when producing a transcript (Braun and Clarke, 2006). What was important was that the transcripts retained the information the researcher needed from the verbal accounts, and in a way which was true to their original nature (Braun and Clarke, 2006). Hence, all the audio-recordings were transcribed exactly as per the original conversations between the researcher and the interviewees.
3.2.7.2. Generating initial codes
The second phase in the data analysis was to extract the phenomena or most significant data from the interviews by assigning conceptual labels, known as codes (Braun and Clarke, 2006).
Interview coding was used to capture what was in the interview data. It helped to move away from particular statements to more abstract interpretations of interview data (Charmaz, 2009). In fact, it has been recommended that the researcher use different coding techniques to examine an interviewee’s responses at different levels (Corbin and Strauss, 2008). In this research, the first coding method was the open coding or line-by-line coding. This method provided a good starting point for the researcher to identify and produce a list of themes of importance to the interviewee.
The coding process began by putting the interview transcripts into a table with three columns:
one for time, one for the full transcript of an interview, and the last for codes. The researcher went through the transcripts, line by line, to write codes for each line manually in the code column. A code or conceptual label was attached to almost every line of the interview transcript to capture what had been said. These labels correspond closely to the interview context. The codes were taken from the interviewee’s own words and the transcripts were read and re-read to carry out further coding and refinement. This process was continuous and entailed comparing
codes from one interview with the codes from a newer interview, which helped to identify prompt questions. After open coding for the first 2 transcripts, the remaining transcripts were coded using the existing codes with new codes added on encountering data that did not fit into existing codes.
Once the line-by-line coding process was completed, the researcher started the process of focused or selective coding which helped the researcher to choose the most telling codes to represent the interviewees' opinions and responses to the questions asked. Focused codes were applied to several paragraphs or lines in transcripts. The researcher used open codes as a starting point to choose the most telling codes to represent the interviewee’s opinion. This process helped the researcher to confirm the adequacy of the initial concepts developed. Once this process was completed, all transcripts and codes were sent to an independent researcher at the School of Healthcare Studies to review and confirm the codes. A discussion was conducted between the researcher and the independent researcher to define and refine the codes and their relationship to each other and to the main question.
3.2.7.3. Searching for themes
The next phase of the data analysis aimed to re-focus the analysis at the broader level of themes, rather than codes. This phase focused on building relationships between codes, themes and different levels of themes. The researcher began the process by analysing the codes and
considering how different codes might combine to form an overarching theme and then sorting the different codes into potential themes, and collating all the relevant coded data extracts within the identified themes.
3.2.7.4. Reviewing themes
Phase four of the data analysis began by developing and refining a set of candidate themes. This phase involved two levels of reviewing and refining the themes identified. Level one involved reviewing at the level of the coded data extracted from the original transcripts. This process involved reading all the collected and extracted codes for each theme and considering whether they appeared to form a coherent pattern. If the themes did appear to form a coherent pattern, then the researcher moved on to the second level of this phase, which was to consider the validity of individual themes in relation to the data set and make sure that the candidate thematic map was
‘accurately’ reflecting the meanings evident in the data set. During this process, the researcher re-read the entire data set to ascertain whether the themes ‘worked’ in relation to it and to code any additional data within themes that had been missed in earlier coding stages if there were any.
However, if the map did not fit the data set, the researcher returned to reviewing and refining the coding again until he devised a satisfactory thematic map.
However, if the candidate themes did not form a coherent pattern, the researcher considered whether the theme itself was problematic, or whether some of the data extracted within it simply did not accurately fit the theme. In that case, the researcher reworked the theme, thus creating a new theme and finding a suitable theme for those codes extracted which did not fit in an existing theme, and/or discarding them from the analysis.
3.2.7.5. Defining and naming themes
Once the “reviewing themes” phase was completed the next phase began. In this phase, the researcher defined and further refined the themes and analysed the data within them. This process included identifying the ‘essence’ of what each theme was about and determining what aspect of the data each theme captured. As part of the refinement, the researcher worked to identify
whether or not a theme contained any sub-themes. Sub-themes are essentially themes within a theme. The subthemes can be useful for giving structure to a particularly large and complex theme, and also for demonstrating the hierarchy of meaning within the data (Braun and Clarke, 2006). By the end of this phase, the researcher clearly defined what the themes were and what they were not.
3.2.7.6. Producing the report
The final phase of the thematic analysis was to produce a report to explain the results in a way which provided a concise, coherent, logical and non-repetitive appraisal of the data. The report was written to provide sufficient evidence of the themes within the data (Braun and Clarke, 2006).