Sample and Process of Data Analysis
4.4 The Process of Analysing the Interviews
4.4.2 Data Reduction, Data Display & Data Interpretation (1) The sub-processes of data reduction and data display (1) resulted in data
interpretation (1), an interpretation that makes extensive use of interviewee
quotations in order to enrich the narrative. This interpretation is based around core themes and can be found in the following chapter (chapter five), with the subsequent discussion of this interpretation being found in chapter six.85
The sub-process data reduction (1) involved coding transcripts of the interviews. As discussed previously, with the exception of one interview, the interviews were recorded. These recordings were transcribed resulting in over 700 pages of text. The coding of the transcripts (data reduction 1) was conducted using a software tool (NVivo). As O’Dwyer (2004) indicates, researchers have mixed views regarding the use of a software tool for coding interviews. A software tool was used in this study because it automates an otherwise manual process. Further there was no concern with the use of software on the part of this researcher, partly because of the newness of the researcher to this type of analysis but also because as O’Dywer (2004) makes clear, software “is merely a tool designed to assist analysis”
(ibid:395). Furthermore, software cannot do the thinking or draw the conclusions for a researcher it is “the individual researcher who has stored and interpreted many of the contextual factors that will influence the inferences drawn from the analysis” (O’Dwyer, 2004: 395).
The process of coding is a systematic method of reviewing and categorising references from transcripts. For this study the process involved perhaps self evidently, reviewing each transcript, identifying pertinent86 text (references) and then categorising (coding) that text under different themes.87 The themes identified
85The reason for this split is because the descriptive release of findings is c15k words long and thus the length of a reasonable chapter in and of itself.
86The use of the term pertinent is telling in this context. The term is being used to indicate that the researcher attempted to identify text that was relevant and that helped to elucidate the interviewees’ points of view, no matter whether or not the reference contradicted a previous reference from the same interview or other interviews.
were informed both from the sub-processes involved in realising data interpretation (0), but also from reviewing the transcript text. In total the process of reviewing and coding the transcripts was conducted three times.
After conducting the coding process a total of 52 themes were identified with 16 of these being major themes and the balance being minor themes (themes underneath a major theme). The major themes had coded text from a minimum of five
interviews (sources), a maximum of 23 interviews and an average of 15.88 That each theme did not contain text from all the interviews is testament to the semi- structured nature of the interviews and the flexibility of this method, where not all of the interviewees were asked all of the questions depending upon the context of a particular interview. Out of the 52 themes in total an interview had text coded to a minimum of 14 themes, a maximum of 28 and an average of 19. The total number of references coded from the transcripts was 894. However as a reference could inform more than one theme there is some double counting in this figure.
Furthermore, on average each transcript had six percent of its text coded, figure that may appear low. However it reflects the nature of conversations, where much of a discussion is filled with linking phrases and re-iterations as opposed to quotable phrases.
Having coded the transcripts, the next sub-process was to display the coded data, data display (1). This display involved two stages, the first was to analyse the themes against the seven interview attributes that were captured89 and the second was to review the coded text and map it to identify linkages and themes that would allow the development of a coherent narrative for data interpretation (1). The first stage reviewing the categorised text against the interview attributes was done via the use of matrices and percentage charts. The matrices were used to facilitate a visual review of the themes against interviewee organisation and whether there were any particular patterns. The percentage charts were used to help identify whether any particular theme appeared to be biased towards one particular attribute, for example are the majority of quotes under a particular theme from female
interviewees or from a disproportionate number of female interviewees relative to
88Each interview had text coded to a minimum of six and a maximum of fourteen major themes with the average being ten.
89The seven attributes are: interviewee gender, organisational size, month of interview, broad area of organisational operations, role of interviewee, whether the interview was face to face or not and the ownership structure of the organisation.
male interviewees. Ultimately this analysis revealed little, however examples of it can be found in Appendix five.
The second stage of analysis involved in data display (1) was the mapping of the themes key messages in order to identify a coherent narrative for data interpretation (1). This was a necessary step because although text can be coded to themes, a collection of themes still requires weaving into a coherent narrative for the reader. A further sub-process within this stage was the identification of contradictions and
divergent messages within the references under a particular theme. After
completing the data reduction and data display, data interpretation (1) was written. This interpretation is a “thick description” (O’Dwyer, 2004:402) of the interview findings and forms the body of chapter five. It uses the themes identified from the coding and makes “extensive use of direct quotations from the transcripts in order to enrich the narrative” (O’Dwyer, 2004:402). Discussion of these findings, the
conclusions and inferences drawn from them and how these findings are reflected in the literature also form part of this interpretation and can be found in chapter six.