Chapter 5: Data and approach to analysis
5.4 Sources of data
There are potential difficulties in using multiple methods resulting in different data sources. Hammersley and Atkinson (2007) warn that it is a mistake to assume that an aggregation of data from different sources will straightforwardly lead to a more complete picture. The different sources did however lead to a richer and more nuanced picture. Observations were conducted in formal settings which to some extent were bound by committee protocols but which were simultaneously idiosyncratic, chaotic and not always particularly ordered within those constraints. Nevertheless, all meetings followed a particular sequence which contained the messiness. The description of these messy and complex meetings is important and I go on in the next chapter to set the scene and evoke the atmosphere of these meetings. However, there were important reasons for behaviour to be circumscribed in the REC. Researchers wanted approval. The REC is the body charged with judging research as ethical or not. In addition, committee members may hold views which they would not express in a meeting. Conducting interviews allowed all participants to express their views and connect to their thinking about key concepts in the study – vulnerability, consent and capacity as well as commenting on REC processes.
5.4.1 Records collected
In ethnography importance is attached to recording. Hammersley and Atkinson (2007) suggested four types of records.
• Condensed accounts which are jottings made as soon as possible after fieldwork, or preferably during it, recording snippets of conversation, routine and significant incidents when and how others reacted.
• Expanded accounts, made when the ethnographer is out of the setting and has more time.
• A fieldwork journal noting the ethnographer’s own feelings and responses and commenting on aspects of the research process.
• Analytic notes record ideas and insights that arise when sorting data, thus beginning the interpretative process.
My own records were not separated out precisely in this way but consisted of a range of condensed notes, lengthier perceptions following the observations and later
the beginning of interpretation when I began to notice patterns in the data collected. There were different styles of notes for observations and interviews.
5.4.2 Records of observations
These correlated to notes made in the field but were not ‘condensed’. Whereas in some settings where the researcher is participant, these notes would of necessity need to be in a shortened form and written opportunistically. Johnson (1997) refers to retreating to the sluice, bathroom or kitchen to make notes whilst working as a bank nurse and researcher to conduct an ethnographic study on a medical ward. I was often sitting (rather than being physically active) at a committee meeting for between 1 – 3 hours. I could be open about recording as everyone knew I was there as a researcher/observer. My notes then were more often lengthy demanding focus and concentration. The audio recording of committee meetings was not permitted by NRES and so notes had to be written at the time of the meeting. The recording of data was adapted from an approach previously used in an observational study of attribution of cause of death in children by staff in an Accident & Emergency department (May-Chahall et al, 2004). This provided a simple framework for recording. Observation data consisted of notes taken at the time of the REC meeting which would be structured and notes taken after the meeting which were unstructured. The notes taken at the meeting consist of a written record of who is speaking, acting: what (doing and saying); the type of interaction or exchange, for example between committee members, to researcher and my own perception of what I think is happening. Unstructured notes taken afterwards consisted of my perceptions of what took place.
Who is speaking What is said Type of interaction/what is happening?
Immediately following meeting – perceptions of what took place
My ‘expanded accounts’ were also a ‘fieldwork journal’ and these notes were made following meetings (usually on the train home). These included my perceptions of what had happened at the observation, anything significant or unusual, differences and similarities between committees and a note of any comments made outside of
the formality of the meeting. These might include coordinators’, Chairs’ or researcher comments. Emotions and feelings about the events of the day were also included. I included here notes about the dynamics of the meeting which I recorded in very informal language, for example, ‘show-off Chairs’ and impressionistic comments on the surroundings for example. These correlated to Hammersley and Atkinson’s ‘fieldwork journal’. The ‘analytic notes’ will be referred to later in the description of how I analysed the data.
In addition to these accounts, I drew small sketches or diagrams when I felt I was able (depending on my proximity to committee members). These simply indicated seating positions of members, where the Chair and Co-ordinator sat and the 'status' (if it was revealed) of committee members, for example, lay members, statisticians, lawyers, nurses, patient representatives and doctors. The diagrams also showed who contributed most in the meeting and between which members most of the dialogue took place.
5.5 Interview schedules
Interviews were conducted as soon as possible after the committee had met but not on the day of the meeting. There were some key outline questions asked of researchers and reviewers but the process of interviews was also organic and followed the direction of the interview.
There were two interview schedules – one for researchers and one for committee members. These were a series of prepared questions, prompts that I adapted according to who was being interviewed and the context. Questions were slightly different for researchers and committee members. Interviews lasted from 30 minutes to one and a half hours. Due to the distances travelled and the location of interviewees, I offered to conduct interviews over the phone or on Skype. Questions to both researchers and reviewers related to process and to consent and its associated requirements. A further question was asked about if they felt anyone was particularly vulnerable in research. Kelly (2013) described a continuum in qualitative interviewing with informality and natural conversation at one end and standardised, structured interviews at the other. I used the schedules to keep the interviews focussed when on the telephone but spontaneous interviews which happened in the field were informal and unstructured. I was conscious that reviewers and researchers
would want to represent themselves in particular ways. I felt that researchers were sometimes initially wary because I might be checking out how sophisticated their understanding was of ethics. However, as the interviews progressed, researchers became engaged in nuanced discussions about their dilemmas. Similarly, my expectation was that reviewers would shape their answers to indicate that they were doing what they thought was required of them. So governed by the terminology of procedure are committee members, that the content of our discussions was often couched in that language. Again, as interviews continued people were likely to talk about uncertainty in the processes and decision- making they were part of. This was assisted by open questions and being open about where the interview might lead. I also slightly changed the focus of questions as I went along. When asking the question: ‘How do you go about making a decision?’ responses usually included comments on the researcher so I extended that question to ask about the attendance of the researcher as this was clearly of importance to reviewers.
Table 8: Interview schedules Committee members
Process questions
• What made you think in the application(s) in the meeting I observed that there should be a favourable or unfavourable outcome?
• What do you take into account when making a decision?
• How do you go about making a decision when you are reviewing? (Extended – presence of the researcher)
• How do you use the procedures in reaching a decision?
• Do you want to say anything about the process, the way the committee works? (Extended – the remit of the REC? lay membership, allocation)
Capacity/consent questions
• What is your understanding of capacity and consent in the context of research ethics?
• Do you think anybody is particularly vulnerable in research? • Do the requirements (MCA2005) and crib sheet assist you?
Researcher questions Process questions
• Could you talk about your experience of attending the REC and the process? (extended – and the application process)?
Capacity/consent questions
• What did you consider in relation to capacity and consent in your study? (extended – previous studies)
• Did the requirements assist you? (either in application or in previous research)
• Do you have particular ideas of who is vulnerable in research? 5.6 Coding
The analytic notes made on sorting the data came about as a result of the transcribing process. Perceptions, observations were made at the time of transcribing both field and interview notes. This process itself constituted a
preliminary analysis and I began to see links in the data and looked at where emerging items were occurring.
Table 9: Data types used in analysis 1 observation notes
2 interview records-committee members including Chairs 3 interview records- researchers
4 field-notes – these included perceptions recorded in field- notes during observation, immediately following and ongoing commentary I kept in journals and then on transcribing.
A number of commentators are prescriptive about when analysis should take place with some indicating that researchers begin the process almost on entering the field and others suggesting delaying analysis until a substantial amount of data has been collected. Bryman and Burgess (1995) describe a developmental position on analysis which begins with coding and progresses to conceptualization. Mason (1996) distinguishes between making data amenable to analysis (or coding) and developing the analysis but points out, that the distinction between the two activities becomes blurred because in establishing some preliminary categories I was involved in analytical and creative thinking. Transcribing was an important activity because it allowed me to immerse myself in the records I had amassed and become familiar with them. In looking across the records, I began to see categories emerging which I classed as nodes in the NVivo system. I felt that I began interpreting when I was transcribing moving from passive reading to ‘active interpretation’ (Silverman, 2013) at this stage. Mason argues that it is important to treat the activities of coding and analysis as distinct in order to emphasize that ‘although techniques like indexing and retrieval provide materials with which an analysis can be created and crafted, they do not represent the analysis in and of themselves.’ (p91).
Coding is heuristic and a step-forward to more rigorous analysis. It is not simply labelling, but leads from the data to an idea and then from the idea back to the data pertaining to that idea (Richard and Morse, 2007). Saldana comments that ‘focused coding’ (Saldana, 2009:155) is useful for ethnographic methods as it encourages the
development of major categories without focusing too much on their properties or dimensions. Coding is described as being a word or phrase which has symbolic resonance. For Saldana, the code may be summative or salient but also be an evocative attribute for a portion of the data collected. He cautions that not all qualitative data are conducive to coding and highlights how methodology and the specific context of research influence the process. Coding is performed through our own interpretive, subjective lens. Finally, he notes that coding is a transitional process which takes place between collecting data and extensive data analysis (p5).
In qualitative analysis of data there is a tension between maintaining a rigorous approach and remaining engaged in the field whilst collecting data, coding data and during analysis. A rigorous approach to data at each stage is linked to validity and reliability in a positivist tradition. Hammersley (1992) argues that validity in qualitative research depends on research being credible and plausible as well as relevant and making a contribution to knowledge and understanding. NVivo may suggest and provide evidence of a systematic approach. It helped to organize the data but is limited in analyses and interpretation as it removes crucial context.
Silverman (2001) proposed some guiding principles in approaching data collection and analysis, which I adapted, and though I did not follow these absolutely, they did guide what I did.
• Firstly, the principle that observational research can be both original and valid and involve testing hypotheses in the field (Silverman, 2001: 69). If there was a feature of the committees which seemed significant to me then this heightened my awareness and I was more aware the next time I attended committee to 'test' this out. For example, I noted early on then when queries were raised, particularly by lay members of the committee, then they were expected to be reassured of their concern by the assertion of medical authority. The medical members of the committee (usually in the majority in the RECs which I observed) were able to refer to their own medical practice and reassure the lay member that what was being proposed was fine. I then developed a hypothesis that professional membership was important. I had
• Comparison. In my initial coding, I paid attention to how often categories had occurred and whether they appeared across sites. I would not want to say that claims made on the interpretation and analysis of data are only valid if occurring in a number of RECs but I have been mindful of how frequently categories occur. I have also looked at similarities and differences of RECs. I was also aware that I compared responses in interviews to how REC members and researchers acted and what they said in the context of the more formal committee meetings.
• Simple tabulations. I did identify references, for example to 'vulnerability', at RECs and then place them in the relevant node. I was then able to count how many times this had occurred in RECs and in interviews. I noted observable (inter)actions and gestures at RECs and considered how frequently they occurred in that context.
• Silverman also highlights the importance of separating out the recording of what is said at meetings and during interviews from interpretation and impressions or sensations (emotional reactions) of the researcher. I did this in a very straightforward way in my notebooks and then in transcribing.
5.6.1 Use of NVivo
I used NVivo to begin ‘coding’ and this was helpful in describing initial categories (or 'nodes' in NVivo terminology. This involved cutting and pasting chunks of transcripts under different nodes, creating them as I went along. The other advantage of NVivo is that each node can be opened to display which record it occurred in and this enabled me to refer back to where specifically (in which type of record) these nodes had occurred. Potentially, a node could occur in multiple records. NVivo also allowed me to count occurrences of references to the nodes overall. Nevertheless, I found there to be some disadvantages. In order to organize data into nodes, it has to be extracted from transcripts. One of the hazards of this is loss of context so that in further developing analysis, it was necessary for me to go back to the original transcript in order to be sure of meaning. For example, if I had extracted a comment and placed it under ‘vulnerability’, I had to go back to look at the context including the type of study under discussion, who was making the comment and where it occurred in the live debate in the committee. Bryman and Burgess (1995) comment on how retaining a sense of context seems to be linked to the researcher’s theoretical
assumptions. It was certainly true that as I was holding ideas in my head, I wanted to check out many aspects of a particular node and the complexity of this preliminary theorizing (checking, rechecking ideas, going back and forth to records to confirm hunches and so on) could not be performed by the software (as far as I was able to use it).
5.6.2 Initial categories
In this stage I was engaged with the records I had generated through data collection in the field. I generated categories as I went through the data and ended with 25 categories. I did not apply these categories, rather I created them extracting comments, examples from the records kept and placing them into categories. The ‘problematics’ delineated in the previous section with which I approached the study and which began to emerge influenced the way I selected data for these categories.
Table 10: Initial ordering categories • Engagement with the study • Design - researcher
• Remit of the committee • Committee expertise or
knowledge
• Considerations of capacity- researcher
• Qualitative and quantitative comments
• Notions of harm
• Confidentiality or anonymity concerns
• Judgement about researcher • Reference to me as observer • Reference to previous studies • Assertions of medical authority
• Concerns with language and communication
• Key points about capacity • Requirements
• Reference to professional status • Perception of committee-
researcher • Vulnerability
• Seeking clarity on requirements • Consent concerns in study • Concern with design • References to paperwork • Committee dynamics
• Concern with 'good' research • Moral and ethical dimensions
5.6.3 Approach to coding
In this section I want to talk about the approach to coding which I adopted. Following Mason (1994) and her distinction between coding and analysis I approached these separately. Mason argued that although the distinctions may become blurred, coding was organising and making data manageable but not analysis in itself. My approach was to code into initial categories (above). This was a laborious task going through transcripts closely and deciding what they represented. Charmaz and Mitchell (2001) discuss the role of grounded theory in ethnography. Their important point, relevant I think to all ethnographic research, is that coding does begin the analysis. Initial coding involves the researcher as this is the stage when one is prompted in ‘taking the data apart and to look at them anew with a theoretical eye.’ (p165, 2004). I was interacting with the data (as well as the participants in the research) and my theoretical position was holding in mind what was taken-for granted in the discourse used in the REC. Codes arose directly from the words or what I thought was tacit in what was being said. I also was conscious of how the context supported the taken- for granted actions and statements or, how the context was relevant, how it impeded or changed actions and statements.
These questions were very much aligned to the methodological approach described by Smith above in that I wanted to look from the particular to make connections with the social. I was keen to look at the ‘everyday’ in review. The assumptions which were so dominant that they were not questioned. I was able to expose this through my interviews with researchers and committee members sometimes – but often had to ask myself the question when reviewing and making this preliminary analysis of