Chapter 2: A methodological framework for researching in real-world settings
2.5 Qualitative data analysis
Qualitative data analysis is appropriate for a research project whereby “the researcher trusts textual data more than numerical data and analyses this data in its textual form instead of transforming it into numbers for analysis, with the objective of understanding the meaning of human action” (p.22, Carrera-Fernandez , et al., 2014). It is complimentary to the NDM paradigm as it helps the researcher to understand psychological phenomena from the perspective of those who have experienced them (Vaismoradi, et al., 2013). Qualitative methods can transform large and voluminous data sets into insightful and meaningful conclusions (Liamputtong, 2009). Such methods can facilitate the data analysis and knowledge representation phases in CTA projects that consist of large and ‘messy’ data sets. Qualitative research has a rich history as some of the earliest psychologists including Wilhelm Wundt (1832-1920), arguably the first ‘psychologist’ (Bringmann, Balance & Evans, 1975), sought to understand human psychology through the subjectivity of people (Marecek, 2003). Yet, with the advent of more mathematical and computerised procedures for quantitative research, it fell out of favour with mainstream psychology (Rennie, Watson & Monteiro, 2002). It is only since the early 1990s that interest in qualitative research has resurged (Carrera-Fernandez, et al., 2014), with more publications in mainstream journals (Madill & Gogh, 2008).
2.5.1 Different qualitative analysis techniques
There are a range of different epistemological perspectives and approaches to conducting qualitative analyses (Vaismoradi, et al., 2013). Madill and Gough (2008) outline how data can be collected from a variety of sources including interview data (e.g. telephone, face to face), collaborative data (e.g. role playing, Delphi groups), naturally-occurring data (e.g. diaries, archival data), observational data (e.g. field notes, recordings) and structured data (e.g. protocols, vignettes). It is most commonly associated with transcribed and textual data from audio- and/or video- recorded interviews (i.e. CDM interviews), focus groups, communications ‘in vivo’ and questionnaires (Reavey, 2011). The general process of analysing textual data involves four steps: (i) initial reading of transcripts; (ii) early coding of transcripts;
49 (iii) refinement of codes (possibly into higher order ‘themes’); and (iv) creating a theoretical argument by linking codes and themes together (Liamputtong, 2009).
Generally, coding is either inductive, whereby the researcher derives codes based on ‘bottom-up’ analyses of the text, or deductive, whereby the text is coded ‘top-down’ in line with prior theories or coding dictionaries (Braun & Clarke, 2006). The type of coding differs depending upon the end-goals of the project. It may involve discursive analysis (i.e. focus on how discourse shapes social/cognitive processing); thematic analysis (i.e. inductive or ‘bottom-up’); structured analysis (i.e. deductive or ‘top-down’); or instrumental analysis (i.e. specific philosophical approach to data) (Madill & Gough, 2008). For example, ‘discourse analysis’ is a discursive method that explores how the language used by individuals interacts with social concepts (Liamputtong, 2009); whereas ‘narrative analysis’ is an instrumental method that focuses on how individuals construct stories that convey meaning to others (Liamputtong, 2009). One of the most popular qualitative analysis techniques is structured ‘content analysis’, whereby the researcher uses a coding dictionary to establish frequency counts of different codes that can be subjected to subsequent statistical analyses (Carrera-Fernandez, et al., 2014; Vaismoradi, et al., 2013). Also common is ‘thematic analysis’, which uses both inductive and deductive processes to provide a rich description of data themes (Braun & Clarke, 2006).
Grounded theory (GT) is similar to thematic analysis, but in addition to describing the data it also aims to generate theoretical conclusions (Glaser & Strauss, 1965; Lo, 2014). As this thesis sought to explore the novel concept of decision inertia in emergency response domains (Table 2.1) and because research on this topic was limited, an inductive GT approach was chosen. GT was developed during the evolution of scientific thought on symbolic interactionism (i.e. a theory of human behaviour that describes how an individual’s sense of self is defined by and altered their social interactions with society) (Annells, 1996) and assumes that humans interact with the world via symbolic interactions, the most obvious way being through language (Sarantakos, 1993). It assumes that a person’s language can be coded to identify their ideas and thoughts about a concept (i.e. the challenges to command), which can contribute to a theoretical model of human interaction with the social world (Amsteus, 2014).
50 The process for conducting GT analyses involves coding textual data (i.e. creating descriptive categories of the data); theoretical sampling (i.e. sourcing data from theoretically relevant sources); and memoing (i.e. keeping an audit of ideas during analysis). This should be performed whilst maintaining awareness of the theoretical sensitivity of conclusions (i.e. thinking about the data in theoretical terms); engaging in constant comparison of the data (i.e. constantly contrasting the data against itself); and identifying the point of saturation (i.e. when to develop a final explanatory theory). It is important that researchers self-monitor their progress during qualitative data analyses to ensure that their conclusions remain grounded in the data (Mueser and Nagel, 2009) and that they remain flexible during coding to adapt to novel findings (Braun & Clarke, 2006; Liamputtong, 2009). GT researchers explore how an individual conveys meaning through language about a certain topic (i.e. challenges to emergency response) and look for commonly occurring patterns between participants that can contribute to theory. Thematic analysis follows a similar inductive process to GT (i.e. transcript, annotate, code, refine codes, organise codes and themes) (Lo, 2014); however, whereas GT aims to create a final theoretical model of the data (Glaser & Strauss, 1965), thematic analysis seeks to provide a detailed descriptive account that does not extend beyond the data set being explored. As the aim of this thesis was to use inductive techniques, but with the fundamental goal of developing theoretical hypotheses that could be tested at RQ2, then GT was deemed the most appropriate method for analysing the interview data.
2.5.2 Limitations of qualitative research
Despite the strengths of qualitative analyses for facilitating an exploratory understanding of real-world data, there are a number of limitations that must be acknowledged. There is a poor distinction between the different types of qualitative analyses, which can cause confusion amongst researchers (Bryant, 2002; Lo, 2014). It is not uncommon to find published papers that mislabel the qualitative methodology that has been used, with ‘grounded theory’ and ‘thematic analysis’ most often confused (Vaismoradi, et al., 2013). Ironically, the celebrated ‘flexibility’ of qualitative methods has inadvertently contributed to the incoherence of these methods as research mix methodological approaches (Holloway & Todres, 2003). As a result, qualitative data is commonly perceived as less scientific than more stringently defined quantitative techniques (Crandall, et al., 2006; Laubschagne,
51 2003). For example, different researchers can generate different conclusions about the same data set depending upon their ability, experience and research focus, which suggests subjectivity and bias in analyses (Glӓser & Laudel, 2009). The variability in research conclusions deviates from traditional positivist approaches to psychology that treat objectivity, validity and reliability of results as benchmarks for research quality (Parker, 2008). Furthermore, as qualitative research is often time consuming and labour intensive, then these contributing negative factors can detract potential researchers from using qualitative methods (Hoffman, 1987).
Yet there has been a growing methodological interest in qualitative analyses over recent years. There has been an exponential increase in the number of qualitative publications in psychology journals; increasing from just 12 papers in 1990 to 529 in 2010 (Carrera-Fernandez, et al., 2014; Madill & Gough, 2008). Interest in the use of interviewing has grown, as traditional quantified techniques tend to reduce behaviour and cognitive processing to numerical values that arguably strip the data of its meaning (Bogner & Menz, 2002). Indeed, when research seeks to establish a rich understanding of the experienced cognitions and true psychology of individuals, such as NDM, it would seem that the “whole is greater than the sum of its parts” (p.108; Crandall, et al., 2006). Parker (2004) suggests that psychology should cut its philosophical ties to ‘science’ and instead draw strength from its flexible approach and ability to facilitate interdisciplinary work. Rather than trying to replicate science, psychology is perhaps better defined as the lynchpin that links science and human behaviour. Harre (2004) goes further to argue that qualitative analyses are in fact more scientific than quantitative methods as they are reflexive, meaningful and specific to the research topic. Furthermore, as researchers are people themselves, it has been questioned whether the claim of true objectivity is even possible, as every researcher will bring along their own ‘deductive’ research qualities that informs their analyses (Malterud, 2001).
2.5.3 Conclusions on qualitative data analyses
Qualitative approaches to data analysis are thus useful for providing rich, contextualised and in-depth conclusions on the psychological experience of individuals. A number of authors have attempted to outline more stringent descriptions and criteria for conducting qualitative research to improve the clarity in this methodological approach (e.g. Liamputtong, 2009; Lo, 2014; Morrow, 2005;
52 Tracy, 2010; Vaismoradi, et al., 2013). Furthermore, a number of recent papers have attempted to clarify the distinction between specific qualitative methods by outlining frameworks to facilitate analysis (e.g. Braun & Clarke, 2006; Carrera-Fernandez, et al., 2014; Liamputtong, 2009). Tracy (2010) outlines eight markers of good quality qualitative research by ensuring research has: (i) a worthy topic; (ii) rich rigor; (iii) sincerity; (iv) credibility; (v) resonance; (vi) a significant contribution; (vii) ethics; and (viii) meaningful coherence. Furthermore, Malterud (2001) defines relevance, validity and reflexivity as essential standards to ensure qualitative research quality. Qualitative conclusions can be bolstered by triangulating research to test conclusions in more quantitative and experimental settings (Madill & Gough, 2008; Parker, 2004). Indeed a mixed methods approach was used in the current thesis by embracing both qualitative CDM analyses (Chapters 4 & 5) coupled with more stringent quantitative assessment of questionnaires following the MTFA simulation (Chapters 6 & 7). The fourth and final section of this chapter will describe how advancements in technology can further enhance qualitative data analyses, via the use of NVivo.