Chapter 3: Methodology and Methods
3.4 Research design
3.4.2 Data analysis
All qualitative data can be manipulated and coded (Schensul, 2008), but the choice of protocol to do so is dependent upon the research question and study design, as well as the data itself. Qualitative research is far from a
‘uniform’ approach, but, as Dornyei (2007, p. 242) points out, nowhere is this diversity more apparent that when it comes to data analysis. Despite this diversity, there are similarities between different forms of qualitative analysis, being: primarily about the analysis of language (text), an iterative process that involves nonlinear, back-and-forth movement between data
analysis and generation, and striking a balance between strict, formalised methodologies and intuitive, fluid analytical positions (Dörnyei, 2007). In any case, data reduction, display and interpretation are required.
For this project, Framework Analysis was chosen as a suitable means of interrogating the generated interview data. Developed by Ritchie & Spencer in the late 1980s, it sits within a broad family of analysis methods called
thematic analysis or qualitative content analysis (Gale, Heath, Cameron, Rashid, & Redwood, 2013), which, although now closely associated with qualitative research began life as a quantitative method of analysis (Dörnyei, 2007, p. 245). Such approaches identify commonalities and differences amongst data, permitting the researcher to draw descriptive and/or explanatory conclusions around themes derived from the data. As Ryan & Bernard (2003) point out, theme identification is one of the most important tasks in qualitative research. A defining feature of the method is the ability to cross reference cases– typically an individual interviewee – with codes–
indicators of meaningful information – in a matrix. This provides a structure for systematically reducing the data. Cases here refer to the units of analysis in the Framework Analysis approach, and not the case of a case study; the interviewees are sub-units of the case study. As an approach to qualitative data analysis, Framework Analysis was ‘designed’ for research projects that have:
1. Specific questions 2. A limited time-frame
3. A pre-defined sample (e.g. those associated with a company, programme or sector of concern)
4. A priori issues (e.g. themes that one can expect to occur as a result of the characteristics of the phenomenon being studied, already agreed- upon definitions and constructs and decisions made in light of existing theory (see G. W. Ryan & Bernard, 2003)
This research project has specific questions, a limited timeframe, sensitivity to certain a priori issues (i.e. those that stem from using AIS as a conceptual framework) and a pre-defined sample (people involved in the UK fresh produce innovation system). Framework Analysis is primarily concerned with analyzing the substantive, common-sense meaning in qualitative data, rather than focussing on the use of language itself (as in discourse analysis, for example) (Ritchie & Lewis, 2003, p. 202). Whilst Framework Analysis may contribute to the generation of theory its primary function is to explain what is happening in a certain situation, particularly where an expected output is improved policy measures (Srivastava & Thomson, 2009). In this respect, Framework Analysis ‘lines-up’ with case study methodology. As such Framework is a suitable option for the reduction, display and interpretation of the data, considering the context of the research questions and aims.
3.4.2.1 Using Framework Analysis
Ritchie & Spencer (2003) describe the process of transforming what is often at first messy, voluminous raw data into a more abstract, analytical form as conceptual scaffolding or “analytic hierarchy”. This process enables the
researcher to make sense of the data and provide an analytical account of what is happening.
In short, familiarisation with the data is used to develop an initial set of themes: this is the ‘framework’ by which subsequent data is categorised (indexing). Once no new information comes forward (i.e. once no new theme emerge) saturation has occurred and data collection can end. Charting is used to find cross cutting themes in the data, which involves creating a
matrix of cases and exemplary thematic codes, by reading across cases and looking for similarities and differences in the framework, enabling the construction of higher-level concepts.
Each of these steps is described in more detail below:
• Familiarisation with the dataset: the researcher should familiarize themselves thoroughly with the data before any further analysis. If the researcher has been involved in transcribing the interview, then this provides an opportunity for early familiarisation. Ritchie & Spencer (2003) consider this the foundation of the analytic hierarchy.
• Identifying initial themes or concepts: the goal here is to establish a
framework or ‘index’, drawing upon recurrent themes in the data and
issues introduced into the interviews though the interview guide (these might be a priori issues). These early themes can then be sorted according to different levels of generality so that the index has
a hierarchy of ‘main’ and ‘subthemes’; they should also stay close to
the data in terms of language and substantive meaning (i.e. themes
should be derived from the data and not superimposed from ‘above’
semantic focus, to descriptive categories that remain close to the data, to more abstract classifications (Ritchie & Lewis, 2003, p. 222).
• Indexing: this involves understanding what is meant by textual data and classifying the whole dataset according to the ‘thematic sets’, or categories, of the index established above. There is more than one way of carrying out this process, but it can be done using ‘computer-
assisted qualitative data analysis software’ (CAQDAS), which ensures that the source of a particular piece of information is not lost. Of course, data is often interlinked, and it is worth noting where these interspersions occur for later analysis; likewise, some data may need to be assigned to more than one category.
• Charting and synthesis: next, it is possible to create a matrix to chart
the main themes (and important associated subthemes – see Table 8 in Chapter 4) against cases (that is, individuals involved in the study). This allows the researcher to read across themes and cases to
develop ‘higher-order’, analytical categorisations of the data. It is
important here to retain the language of the respondents without quoting data verbatim (Ritchie & Lewis, 2003).
Once these steps have been taken with the whole dataset, it is possible to begin more a thorough process of developing explanations for accounts by reading across the synthesised data; Framework Analysis permits the rather
rapid appraisal “up and down” the analytical hierarchy to make links between different concepts (Ritchie & Lewis, 2003, p. 256). However, developing full explanations for observed phenomena requires the researcher to also draw upon exiting literature and other theoretical frameworks to explain what they have found (see Discussion).
3.5 Concluding remarks
In this chapter, a methodological outline for answering the research questions has been provided and justified. This began with an exploration of the most fitting analytical framework through which to guide the study, including an explanation of the implications adopting a systems approach, as well as defining the boundaries of this system, before moving onto the case study method and data requirements of this approach. It then outlined the process of conducting semi-structured interviews, including the ethical considerations of this type of study, before discussing how this data can be analysed using the Framework Analysis approach. As such, the methodology and methods used in this thesis can be summarised as follows:
❖ Conceptual framework: Agricultural Innovation Systems
➢ Methodology: embedded, single-case study
▪ Methods:
• Semi-structured interviews with industry practitioners chosen on the basis of:
Their role
Their location
Their position within business or organisation
• Data analysis: Framework Analysis
The effectiveness of this approach, as well as its limitations, is discussed in in Section 5.4.