3.3 Data Collection and Analysis
3.3.4 Analysis of Interview Data
3.3.4.1 Transcription
As part of the informed consent process, permission was sought from participants to digitally record interviews. All digital recordings were then anonymised by me and then transcribed either by me, or by a professional transcription service that adhered to MRC guidelines on confidentiality. Each participant, along with people they described and recorded as part of the ‘map’
of their ‘personal community’, was assigned a pseudonym, or had their name removed, to protect their privacy6. Other identifying information, such as the person’s place of work, hometown, and employment status, etc., were removed or changed in the interview transcript to reduce the possibility of deductive disclosure.
Liamputtong (2013) suggests that where possible transcription should be carried out be the person who conducted the interview, as this can enable researchers to learn much about their own skills as an interviewer. Furthermore,
Liamputtong echoes Kvale’s assertion that transcribing the interview can also
6 For each of the first six interviews, pseudonyms were assigned to each person on the
participant’s ‘affective’ map. Thereafter, given that a large number of people were often included on each map, names were removed/omitted and identifying titles such as ‘sister 1’ or ‘uncle 2’
used instead. The exception to this was where there were a very small number of people included on a participants map. See Gary and Theo’s maps in Appendix 19 as examples.
serve to “have the social and emotional aspects of the interview situation present or reawakened during transcription” (Kvale, 2007, p. 95), thus allowing a researcher to begin the analysis (through transcription) of the meaning of what participants said. As time constraints meant that interviews were transcribed by a professional transcription service, I attempted to follow the principles of this by listening to audio files (with the transcript) jotting down notes, memos, and further reflections on the interview. I was able to link these to the field notes made after interviews.
3.3.4.2 Thematic Analysis
A thematic approach to data analysis was used. In line with the work of Spencer and Pahl (2006), the Framework approach was used to manage data and
facilitate analysis of the substantive content of the interview data (Ritchie and Spencer, 1994; Ritchie et al., 2003b; Ritchie et al., 2003c; Spencer and Pahl, 2006). Framework was originally developed by researchers at NatCen as a method for systematic data management and analysis within applied policy research (Ritchie and Spencer, 1994). Framework offers a systematic approach to thematic analysis, enabling the identification of clear stages/processes that link stages of interpretation. Using Framework enables researchers to develop descriptive accounts by synthesising key categories and presenting them in matrices. Ritchie and Spencer (1994) outline five stages in involved in the
process of data analysis; familiarisation, identification of a thematic framework, indexing, charting, and mapping and interpretation. This approach is well suited to identifying patterns across the data, and enabling comparisons within and across cases, enhancing interpretation and the development of analytical
explanations. Thus, using Framework enabled me as researcher to move beyond a descriptive account, to provide explanations based on interpretation grounded in the data (Spencer and Pahl, 2006). In the sections that follow, I provide an explanation of how I approached the steps outlined by Ritchie and Spencer.
3.3.4.3 Use of NVivo and the ‘Framework’ Approach in Analysis of Interview Data
Data were coded thematically (Braun and Clarke, 2006). I first began the process of familiarising myself with the interview data by reading and re-reading each transcript, jotting notes and memos, and referring back to any fieldnotes. My
aim at this point was to develop a broad thematic coding framework that would facilitate further analysis and enable me to answer my research questions.
During this initial process, my supervisors read a selection of the transcripts, commented on the initial coding, and we discussed emergent themes.
Each of the interview transcripts and participant maps were uploaded to NVivo 9 to assist with data management and organisation. Each participant was
designated as a ‘case node’, and information from the brief questionnaire, along with information about recruitment route, was used to define these ‘cases’. This later facilitated my exploration of patterns across the data. The broad codes developed in the initial stage of familiarisation were used as a basis for coding using the qualitative data analysis software. I coded all text representing
participants ‘talk’ around each theme into one ‘theme node’ within NVivo. This enabled retrieval of the coded data, and further coding within this overall theme. The initial themes coded were; family and families, friends and friendship, partners/boyfriends/sexual partners, identity, ‘community’
(general), ‘gay community/ies’, support (general), support and advice around sex and sexual health, sex and sexual practice, safer sex, HIV and HIV risk/risk management, condoms and condom use. Once the initial coding framework had been finalised I returned to all of the interview transcripts to ensure that all interview data were coded consistently in line with the coding framework.
Given the large amount of textual data generated, breaking down the data into these broad themes enabled me to return to the data in more ‘manageable chunks’. Using NVivo meant that it was also easy to return to the transcript as a whole, enabling me to examine specific sections of the text in the context of the wider interview, thereby facilitating movement between sections of data and the interview as a whole. Although analysis can be conducted within NVivo 9 software, at this stage I developed a series of framework templates for some of the broad thematic codes. For example, I developed frameworks for condom use, support and advice, support and advice around sex, ‘communities’, and created separate matrices for HIV testing and perceptions of HIV risk to self and others. In the frameworks I made notes and summarised key points as well as including key quotes from the transcripts. Where new issues or themes emerged from my engagement with the data, I then reviewed coded transcripts to ensure
I had not missed relevant data, thereby engaging iteratively in a process of analysis. Using NVivo and Framework in this way also drew attention to cases where data did not fit an overall pattern observed. For example, as part of the interview men discussed specific sexual practices - one participant had a very different perspective on oral sex than the others - and systematic coding and charting meant that this was easily identifiable. Using this systematic approach enabled me to consider similarities and differences between his perspective and that of other participants. A small extract from a framework is included below.
Figure 3-1 Extract from Framework Template
Having ‘charted’ the data in this way I was able to move between the
frameworks created, coded themes in NVivo, and the transcripts as a whole to examine patterns across the data. Movement across the data in this way facilitated interpretation of the data by assisting me in identifying patterns, specifically similarities and differences across the data. I could then begin to develop explanations for these. This process corresponds to the ‘mapping’ and
‘interpretation’ described by Ritchie and Spencer (1994). Given the breadth and depth of the data collected, it is not possible to present all of the findings from my analysis within this thesis. With this in mind, the thesis is structured in such a way as to present themes which help answer my research questions.
3.3.4.4 Integrating Men’s Affective Maps into Analysis
In addition to using the affective maps as a means of exploring men’s personal relationships during interviews, I was able to use them as part of the process of analysis to explore patterns across the men’s ‘personal communities’. Examples of some of the anonymised maps are included in the chapter that follows, and a copy of each of the coded maps can be found in Appendix 17.
Given that distinctions are made between different groups – friends, family, colleagues, and professionals – within both the literature and in the men’s descriptions of their ‘personal communities’, each of the maps was coded according to who the men had chosen to include. As far as possible such
classifications were based on participants’ definitions of the different individuals (and groups) they chose to include.
Figure 3-2 Coding Applied to 'Affective Maps' Developed by Participants
Based on the coding applied, each of the men’s ‘personal communities’ were categorised on three levels: whether they were ‘friend dominated’ or ‘family dominated’; whether their friendship groups were predominantly with other gay (LGB), or straight friends; and whether friendship groups were patterned in terms of gender. This is explored further in the first findings chapter.