See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/278961843 Qualitative Data Analysis and Interpretation: Systematic Search for Meaning Research · June 2015 DOI: 10.13140/RG.2.1.1375.7608 CITATIONS READS 64 107,267 1 author: Patrick Ngulube University of South Africa 204 PUBLICATIONS 2,804 CITATIONS SEE PROFILE Some of the authors of this publication are also working on these related projects: Research Data Management in sub-Saharan Africa View project Records management and public service delivery in Kenya View project All content following this page was uploaded by Patrick Ngulube on 22 June 2015. The user has requested enhancement of the downloaded file. Chapter 81 Qualitative Data Analysis and Interpretation: Systematic Search for Meaning Patrick Ngulube 8.1 Introduction In this chapter, we will discuss the analysis and interpretation of qualitative data as a kind of follow through on Chapter 7 (seven) discussions. The approaches to qualitative and quantitative data analysis are different, as illustrated in table 8.1 below. The remarkable growth of qualitative research in many disciplines, including business and management (Myers, 2009), health and social sciences (Atkinson, Coffey & Delamont, 2001; Flick, 2002), and psychology (Madill & Gough, 2008), makes it imperative for researchers to be familiar with qualitative data analysis. An understanding of qualitative data analysis is fundamental to their “systematic search for meaning” (Hatch, 2002:148) in their data. Qualitative data analysis in one of the most important steps in the qualitative research process (Leech & Onwuegbuzie, 2007) because it assists researchers to make sense of their qualitative data. The process of qualitative data analysis is “labour intensive and time consuming” (Lofland, Snow, Anderson & Lofland, and 2006:196). This is partly due to the fact that qualitative research produces “large amounts of contextually laden, subjective, and richly detailed data” (Byrne, 2001:904). The “true test of a competent qualitative researcher comes in the analysis of the data” (Henning, Van Rensburg & Smit, 2004:101). Qualitative data analysis is concerned with transforming raw data by searching, evaluating, recognising, coding, mapping, exploring and describing patterns, trends, themes and categories in the raw data, in order to interpret them and provide their underlying meanings. Patton (2002:41) refers to this process as inductive analysis and creative synthesis. After reading this chapter, you should be able to: Identify the main source of qualitative data; Differentiate between qualitative and quantitative data analysis; Identify various approaches to analysing and interpreting qualitative data; Describe key qualitative data analysis procedures; Explore computer-based qualitative data analysis procedures; Outline data management procedures; Discuss common criteria for evaluating qualitative data; and Reflect on qualitative data interpretation. 1 How to cite this chapter: Ngulube, P. 2015. Qualitative data analysis and interpretation: systematic search for meaning, in Mathipa, ER & Gumbo, MT. (eds). Addressing research challenges: making headway for developing researchers. Mosala-MASEDI Publishers & Booksellers cc: Noordywk, pp. 131-156. 1 This chapter examines various strategies for qualitative data analysis and interpretation. Furthermore, it discusses quality and rigour in qualitative data analysis. It is apparent from the list of references in this chapter that much has been written on qualitative data analysis. The proliferation of literature has certainly contributed to the complexity of the approach towards qualitative data analysis. Birks (2014:224-225) writes that: The complexities of qualitative research terminology and some of the original work on the topic can leave a novice researcher a little overwhelmed when commencing a study (and often throughout the later stages). However, little guidance is provided on how the various types of qualitative data analysis relate to a qualitative research project. It could be that the analysis of qualitative data is still “a mysterious, half-formulated art”, as Miles asserted in 1979. Consequently, the issues related to the analysis of qualitative data are still under the spotlight three decades later. In fact, many qualitative researchers mistakenly believe that the only way to analyse qualitative data is by means of constant comparative or constant comparison analysis, as suggested by Glaser and Strauss (1967) (Leech & Onwuegbuzie, 2007:562). We concede that it is not always possible to separate data analysis from data collection in qualitative studies, as analysis sometimes occurs during data collection, but we are deliberately making this phase distinct in this chapter, because most of the analysis of the qualitative data tends to take place at the end. It is at the end that answers are sought to questions such as: What does this data mean? What are the major themes emerging from the data? Does the data contribute to a further understanding of the field? There are a variety of approaches used to analyse qualitative data. Consequently, this chapter provides a few illustrative examples using the framework described in section 8.5. Researchers should take note of the fact that whichever approach to analysing qualitative data is adopted, the data analysis procedure should be aligned to the data that has been gathered and the assumptions of the research approaches. Researchers should also note that “all forms of qualitative data analysis involve interpretation and the researcher must always acknowledge the possibility that alternative interpretations are possible” (Harding, 2013:139). 8.2 Differentiating features of quantitative and qualitative data analysis Table 8.1 below illustrates the differences and similarities in data analysis between the qualitative and quantitative methodologies. Although data analysis in qualitative and quantitative traditions is based on different assumptions, data analysis transcends the qualitative-quantitative divide, which is partially discussed in Chapter Nine. Data analysis in both research traditions is not a series of binary oppositions. It is possible for data analysis to have moments of qualitative and quantitative approaches. This partly explains why some qualitative data analysis software programs such as Atlas/ti easily interface with SPSS©, a statistical package. It is unhelpful and unproductive to view data analysis in the two research traditions as mutually exclusive. 2 Table 8.1: Similarities and differences between quantitative and qualitative data analysis Quantitative Qualitative Collecting and analysing data straightforward and not stressful usually Helpful in “answering questions of who, where, how many, how much, and what is the relationship between specific variables” (Adler, 1996: 5) Hard data are collected, as they are in the form of numbers, counts and other statistical formulae Clear and formulated conventions for data analysis and process is predictable Quantitative research produces narratives which document the course of the project and give an audit trail of the research (Boulton and Hammersley, 2006::245) Data analysis is usually done at the end when all data has been collected in a linear fashion Not flexible and is usually difficult to followup on promising hunches Standardised data is collected through measuring either qualitative or quantitative variables The researcher seeks to verify or test a theory and the approach tends to be confirmatory Relationships between independent and dependent variables is of major concern (tends to be variable-centric) Biased towards hypothesis testing and deductive-oriented Use of computer programs and software to analyse data is possible (e.g. SPSS©, linear structural relations (LISREL) and Statistical Analysis System (SAS)) Collecting and analysing data is highly labour-intensive and generates a lot of stress (Miles, 1979). For instance, Adair and Pastori (2011) had 150 focus group interviews in multiple languages that they had to analyse Provides naturally occurring information and assists in answering why and how questions, while documenting the interventions of the researcher during the whole research process Soft data are collected, as they are in the form of words (texts, images, artefacts, narratives) and everything else (Blaxter, Hughes and Tight, 2006) Methods of data analysis are not clearly formulated (Miles, 1979:590) and process is not predetermined (Suter, 2012:343). Qualitative research produces a sequence of events independent of the researcher (Boulton and Hammersley, 2006::245) Data is analysed as they are collected (Glaser and Strauss, 1967; Miles and Huberman, 1994) because data “collection and analysis are interactive and occur in overlapping cycles” (McMillan and Schumacher, 2014:364) Flexible enough to make adjustments during data collection, as supplementary questions may be formulated during data collection to gather additional data Huge amounts of data are collected that need to be summarised and interpreted The researcher puts “aside perceived notions about what the researcher expects to find in the research, and letting the data and the interpretation of it, guide analysis” (Corbin and Strauss, 2008:160) and the approach tends to be exploratory Focus on the meaning of events and actions as expressed by the participants (case-centric) (Plowright, 2011) Favours analytical induction Use of computer programs and software to analyse data is possible (e.g. Nvivo, previously NUD*IST and Atlas/ti) It is clear from the above table that qualitative data is generally analysed from the beginning of the research, as suggested by Miles and Huberman (1994:50). The analysis of the collected data assists the researcher to devise strategies to generate more data, in order to answer the research question. The focus of qualitative analysis is on the meaning of events and actions, rather than statistical significance and relationships between variables. Table 8.1 draws an interesting comparison to table 12.1 in Suter (2012). 8.3 Sources of qualitative data Qualitative research uses all sorts of data (Braun & Clarke, 2013). Depending on the research questions informing a study, qualitative empirical materials may be 3 obtained through the utilisation of qualitative designs or approaches, such as the case study (situated knowledge), historical research (knowledge of history), grounded theory (knowledge of process and outcome), ethnography (knowledge of culture), content analysis (knowledge of content), phenomenology (knowledge of lived experience), action research (knowledge of process, outcome and change), hermeneutics (knowledge and interpretation of the scriptures or text) and discourse analysis (knowledge of discourse) (Mills, 2014:35). This is not an exhaustive list, as other approaches can be found in major texts on qualitative methods. Each of these designs has different purposes and prospective outcomes. Many research methods texts confuse research designs with methods. According to De Vaus (2001:9), “It is not uncommon to see research design treated as a mode of data collection rather than as a logical structure of the inquiry”. For instance, Payne and Payne (2004:175) refer to research designs as research methods. This resonates with Mills’ (2014:36) conceptualisation of research methods as including data generation and collection, analysis of data, quality and rigour, and the interpretation of findings. However, Creswell (2013:5) uses the term “research methods” to refer to techniques such as questionnaires; interviews; observation; document analysis; and artefact analysis. With reference to Rule and John (2011) and Creswell (2013), we use the term research methods to refer to techniques for gathering data, while research designs or research approaches are ways of designing and conducting research. Qualitative research designs or approaches are as diverse as sources of qualitative data. The major sources of qualitative data may be observations, interviews, questionnaires, physical traces, document review and audio-visual materials (Patton, 2002; McMillan & Schumacher, 2014). However, most qualitative research mainly relies on interview data (Perakyla & Ruusuvuori, 2011). This is also evident in disciplines such as psychology (Madill & Gough, 2008) and education (Leech & Onwuegbuzie, 2008). 8.4 Qualitative data management Qualitative data are gathered and constructed from relatively few sources, but the amount of data generated tends to be extensive. A structured mechanism for managing research data contributes to the credibility of the research outcome (Birks, 2014). The way in which qualitative data and resources are managed contributes to procedural precision and the preservation of the quality of the research (Birks & Mills, 2011). However, there is no widely accepted system of recording qualitative data (Williamson, Given & Scifleet, 2013), but it is clear that some system is necessary (Lofland et al., 2006). The major factor that should determine the researcher’s choice is the logic and security that the system provides (Birks, 2014). Electronic files are useful in storing transcribed interviews, observation data and memos. Asking the following three questions suggested by Miles and Huberman (1994:46) will assist a novice researcher in managing research data using computers: What will my files look like? How will they be organised? How can I get the information from them that I need? 4 The files must be backed up, irrespective of what system is utilised. Printed copies may be necessary when analysing data, as it is easier to immerse oneself in one’s data using hard copies than electronic copies (Williamson et al., 2013). 8.5 Creating a picture from pieces of gathered data Creating meaning and making sense of the data is the main purpose of qualitative data analysis. Miles and Huberman (1994:10) noted that “the strengths of qualitative data rest on the competence with which their analysis is carried out”. The methods of data analysis are based on three qualitative data analysis strategies identified by Creswell (2007), including preparing and organising the data, coding, and presenting the data in the form of text, tables or figures. There are various types of qualitative data analysis and their utilisation depends on within which framework qualitative research was adopted. Research questions are used as a guide for conducting the analysis, as for instance, each question becoming a major coding category that is broken down into sub-categories. Although not all qualitative data analysis is inductive (Madill & Gough, 2008), the inductive analytical process is a common characteristic of qualitative data analysis (Curtis & Curtis, 2011). The common denominator for the data analysis procedures described in this chapter is that they all involve data reduction, data display, and conclusion drawing verification (Miles & Huberman, 1994). After having demonstrated how little focus is placed on qualitative data analysis, Leech and Onwuegbuzie (2007:563) described seven commonly used techniques for analysing qualitative data – method of constant comparison, keywords-in-context, word count, classical content analysis, domain analysis, taxonomic analysis, and componential analysis. Later on, they expanded the list to eighteen (Leech & Onwuegbuzie, 2008:588). Dawson (2009:119-125) divided qualitative data analysis into four components: thematic analysis, comparative analysis, content analysis and discourse analysis. On the other hand, Madill and Gough (2008:257) categorised methods of analysing qualitative data as discursive, thematic, structured and instrumental. Drawing in part from Madill and Gough (2008:257), we have thus framed our discussion of the different ways of doing qualitative data analysis into these four groups. These categories are not mutually exclusive and it is possible to combine some of these categories to illuminate the understanding of the phenomenon under investigation and to achieve data analysis triangulation, as conceptualised by Leech and Onwuegbuzie (2007). Although the list of Madill and Gough (2008) may not be exhaustive, it is sufficiently comprehensive and offers a helpful framework for conceptualising and understanding qualitative data analysis. Furthermore, a closer look at the various data analysis techniques outlined by Leech and Onwuegbuzie (2007; 2008) reveals a partial overlap with the categories outlined by Madill and Gough (2008), even though they are articulated in other terms. The categories that were outlined by Dawson (2009) partly cover the components of the typology of Madill and Gough (2008). For the purpose of this chapter, one example of the data analysis procedure under each category described by Madill and Gough (2008) is provided. The reader is encouraged to further investigate the repertoire of data analysis procedures that are mentioned in various subsections of this section. 5 Qualitative data analysis is at times guided by theory and theoretical concepts, as explained in Chapter Four, but is “always shaped to some extent by the researcher’s standpoint, disciplinary knowledge and epistemology” (Braun & Clarke, 2013:175). Key questions that should inform qualitative data analysis and be asked continuously during the process have been outlined by Hair, Jr and others (2011:282): What themes and common patterns are emerging that relate to the research objectives? How are these themes and patterns related to the focus of the research? Are there examples of responses that are inconsistent with the typical patterns and themes? Can these inconsistencies be explained or perhaps used to expand or redirect the research? Do the patterns or themes indicate that additional data, perhaps in a new area, needs to be collected? (If yes, then proceed to collect the data). Earlier on, Henning, Van Rensburg and Smit (2004:106) pointed out that similar questions may be asked after collected data has been coded and categorised, in order for the qualitative researcher to see “the whole”. 8.5.1 Coding According to Leech and Onwuegbuzie (2007:565), coding is a term used to refer to the method of constant comparison analysis, which was conceptualised by the fathers of grounded theory, Glaser and Strauss (1967). However, coding can be used in other qualitative approaches, independently from grounded theory (Miles & Huberman, 1994; Patton, 2002). Coding plays a key role in category identification in qualitative data analysis (Williamson et al., 2013). It is a process that assists the researcher to move data to a higher level of abstraction (Ng & Hase, 2008:159). The use of codes was further reinforced by the increased use of computer management programs in qualitative research (Grbich, 2013). The ability of computer programs to process qualitative data mainly depends on how the data is codified into words, phrases, sentences or paragraphs. The aim of coding is to “break down and understand a text and to attach and develop categories and put them into an order in the course of time” (Flick, 2002:178). The coding process involves the grouping and labelling of segments of data. It also assists in identifying and connecting bits of data. Miles and Huberman (1994:56) state that: Codes are tags or labels for assigning units of analysis to the descriptive or inferential information compiled during a study. Codes are attached to “chunks” of varying size – words, phrases, sentences, or whole paragraphs, connected or unconnected to a specific setting. This definition seems to suggest that although there are various qualitative research approaches, there is not much variation when it comes to the process of coding data or what Miles and Huberman (1994) refer to as data reduction. The fact that coding is one of the most commonly used qualitative data analysis methods may have contributed to the misconception by many qualitative researchers that constant comparison analysis or coding is the only data analysis technique (see Leech & Onwuegbuzie, 2007). 6 Data analysis in various qualitative research approaches begins with coding (Saldaña, 2009; Liamputtong, 2013). The stages of data analysis are similar, as the process is iterative (i.e. moving backwards and forwards), revolving around the research questions or theoretical frameworks identified from the literature and reducing the data into segments and groupings, which are finally linked to the literature and theory as data are interpreted. Figure 8.1 below gives examples of some of the qualitative designs which use coding as the initial step in analysing data. Critical ethnograpghy Action research Content analysis Discourse analysis Grounded Theory Phenomenological analysis Data coding and constant comparison Some qualitative data collection approaches Figure 8.1: Examples of qualitative research designs that employ coding There are a number of coding options available to researchers. Grbich (2013) suggested the following options: Coding and developing themes or major codes manually or using a computer program; Developing themes through thematic analysis and then coding the data around the themes; Summarising and presenting data with minimal coding; and Using research questions to develop broad themes. The last two coding options apply mostly to the phenomenology and autoethnography research designs. According to Boyatzis (1998: x-xi), elements of a good code include the following: Labels; Definitions of what each theme concerns; Descriptions of how to know when the theme occurs; Descriptions of any qualification or exclusions to identifying themes; and Examples to eliminate possible confusion when looking for themes. These elements assist researchers to maintain an audit trail, in order to demonstrate the trustworthiness and credibility of a study, as described in section 8.8. Byrman (2012) provides a useful list of coding steps that a researcher may use as a starting point (see figure 8.2 below). 7 Commence coding while collecting data Figure 8.2: Coding steps (Adapted from Bryman, 2012:575-577; see also Saldaña, 2009) The arrows in figure 8.2 above are facing in both directions, in order to highlight the fact that the process of coding is recursive or iterative. Codes are usually revised and refined during analysis in a backwards and forwards fashion. Readers should note that it is possible to use codes in more than one category. The researcher constantly compares the categories in search of meaning. It is noteworthy that categories in grounded theory are developed from data, in contrast to categories in content research or citation analysis, where they are at times predetermined or “brought to the empirical material” (Flick, 2002:190). Content research is discussed in greater detail in section 8.5.4.1. In a nutshell, the major tasks associated with coding include sampling, identifying themes, building codebooks, marking texts, constructing models and testing the models (Ryan & Bernard, 2000). Sampling involves identifying texts that are analysed; identifying themes, which entails deriving themes from the data or the literature; building codebooks, which include a listing of codes and their definitions; marking texts, which refers to assigning codes to units of text; construction of models, which establishes the link between themes and concepts; and testing of models constructed in the previous step, which is the final task in coding. At times, qualitative researchers use codes and themes interchangeably. This tendency is understandable if one takes into consideration the fact that qualitative data analysis mainly involved the identification of dominant themes or thematic analysis, before Glaser and Strauss (1967) introduced the concept of coding. However, Grbich (2013:259) cautions researchers to be “transparent about how they are using such labels in the data analytic process”. Thematic analysis may be undertaken without coding. Some researchers use codes to develop themes and 8 others start with themes in order to come up with codes. On the other hand, some researchers use only themes or only codes. 8.5.2 Memos Glaser (1978) points out that: memos are fundamental to doing grounded theory, because this makes the researcher stop and analyse the data and codes from the start of the research. However, memos may be used in various qualitative contexts (Harding, 2013:110). Memos can serve as an audit trail of the research process. They give details of what the researcher was doing during the process. Memos enable researchers to step back from the data and move beyond codes as they think aloud reflectively and conceptually (Miles & Huberman, 1994:72). Writing memos assists the researcher to move directly into an analysis of the data and to systematically examine, explore and elaborate on bits of data and early codes (Charmaz, 1990:1169). The main types of memos include procedural memos and analytic memos (Myers, 2009). Procedural memos provide a trail of the research process. They help a researcher to be accountable and transparent during the research process, as described in section 8.8. Analytical memos are the researcher’s commentary on what the data may mean, as they are notes added to coded segments. They are tools for developing concepts and themes. Analytical memos may be categorised as code memos and theoretical memos. Code memos are utilised in open coding, while theoretical memos are used in axial and selective coding. 8.5.3 Thematic data analysis Thematic analysis is “possibly the most widely used method of data analysis, but not “branded” as a specific method until recently” (Braun & Clarke, 2013:175). Thematic data analysis procedures are related to qualitative methods such as grounded theory, framework analysis, interpretative phenomenological analysis, critical ethnography and template analysis (Madill & Gough, 2008). Thematic analysis is considered to be the foundational approach to qualitative data analysis (Braun & Clarke, 2006; Williamson et al., 2013). As explained in section 8.5.1, coding was only introduced as a concept in qualitative data analysis in the 1960s. Thematic analysis is “A method for identifying themes and patterns of meaning across a dataset in relation to a research question...” (Braun & Clarke, 2013:175). Flick (2002) identifies two types of thematic analysis, namely theoretical coding and thematic coding. Theoretical coding was developed by Glaser and Strauss (1967) to analyse data gathered for developing a grounded theory. Thematic coding differs from theoretical coding, although it is premised on the same assumptions. The process starts with specific data that is then transformed into categories and themes. The conclusions are drawn based on observations from the transformed data. The focal point of thematic analysis is category coding. Byrne (2001:904) suggests that thematic analysis is analogous to sorting a box of buttons by grouping them according to “size, number of holes, color, or type”. 8.5.3.1 Thematic data analysis as an example of thematic coding Few texts provide guidelines on how themes are identified (Grbich, 2013). In contrast to grounded theory, thematic analysis does not include theoretical sampling 9 (Liamputtong, 2013). Repetition of terms and typologies may assist in generating analytic patterns or themes (Braun & Clarke, 2006:86). The aim of thematic analysis is to generate thematic domains, in contrast to developing core categories, as in theoretical coding. A case is analysed to determine themes and the themes that emerge are used to analyse further cases in a comparative fashion. Flick (2002:211) states that the focus of thematic analysis is “on conducting case studies and only at a later stage is attention turned to comparing and contrasting cases”. Open and selective coding may be used to analyse the first case, and the themes that emerge are used as a basis for comparison with further cases. The steps in thematic analysis are outlined in figure 3 below. Transcribe Take note of items of interest Code across the entire data set Search for themes Review themes by mapping provisional themes and their relationships Define and name themes Finalise analysis Figure 8.3: Steps in thematic analysis (Adapted from Braun and Clarke, 2013; 2006) However, Harding (2013:112) suggests four steps that are involved in analysing themes: Identifying the theme and creating a category; Collating codes from different illustrative issues into the category; Creating sub-categories to reflect different elements of the themes; and Utilising the themes to explain relationships between different parts of the data and building theory. Although the steps in analysing themes differ in the two examples given above, it is clear that there is an overlap between the steps provided in these examples. 8.5.3.2 Grounded theory data analysis as an example of theoretical coding Regardless of which grounded theorist’s approach is adopted, grounded theory data analysis comprises three phases of coding, as illustrated in table 2 below. Both the Glaserian and Straussian grounded theorists agree that the aim of grounded theory is to “... generate core concepts and develop a theoretical framework that specifies their interrelationships” (Parker & Roffey, 1997:222). The data analysis phases include the following (Mills, Birks & Hoare, 2014): 10 initial coding; intermediate coding; and advanced coding. Holton (2007) refers to the first two phases as substantive coding. On the other hand, Charmaz (1995) collapses the last two phases of coding into what may be loosely termed focused coding. Tan (2010: 102) states that: “Glaser and Strauss (1967) originally did not clearly name the data analysis process as open coding or theoretical coding, but emphasised the constant comparative method for generating theory”. “Substantive coding”, which is comparing incident to incident to generate categories and comparing new incidents to these categories; “theoretical coding”, which is conceptualising how the substantive codes may relate to each other as hypotheses to be integrated into a theory; and “coding families” as the analyst’s coding procedure, were only introduced by Glaser (1978:72) in later works. The traditional and evolved grounded theorists denote the initial phase of coding as open coding, while the constructivists prefer the term initial coding. The traditional, evolved and constructivist grounded theorists regard intermediate coding as selective, axial and focused coding respectively. Advanced coding, which is the third and last phase in the process of data analysis, is known as theoretical coding by the traditional and constructivist grounded theorists, but the evolved grounded theorists view it as selective coding. Table 8.2 summarises the conceptualisation of these phases of data coding by the various grounded theorists. Table 8.2: Phases of grounded theory coding Genres of Grounded Theory Initial Intermediate Advanced Traditional (Glaser and Strauss, 1967) Evolved (Corbin and Strauss, 2008) Constructivist (Charmaz, 2006) Open coding Open coding Initial coding Selective coding Axial coding Focused coding Theoretical coding Selective coding Theoretical coding Adapted from Birks and Mills (2011) and Mills, Birks and Hoare (2014) Flick (2002) cautions that the phases should not be treated as distinct, as they are just different ways of dealing with textual sources. The phases are used to handle data in a linear or iterative manner. Ultimately, “data are broken down, conceptualised, and put together in new ways” (Strauss & Corbin, 1990:57). The researcher may move back and forth between the phases as he or she interprets the data. The process of data analysis involves constant comparison of the differences and similarities in the data, in order to come up with themes and patterns. Theoretical saturation results from the comparison of incidents or indicators. Theoretical saturation is reached when new codes are not being generated. a) Initial coding (descriptive codes) The initial phase of coding that transforms data into codes is known as open coding or initial coding. It is the “initial step of theoretical analysis that pertains to the initial discovery of categories and their properties” (Glaser, 1992:39). Journal notes, interviews and observations are broken down into phrases and keywords. Open coding aims at describing the overall features of the data “by breaking down, analysing, comparing and categorising the data” (Eriksson & Kovalainen, 2008:16011 161). It expresses data in the form of concepts (Flick, 2002:177). As discussed in Chapter Four, concepts are the building blocks of theory. Strauss and Corbin (1990) reiterate this point when defining open coding: Concepts are the building blocks of theory. Open coding in grounded theory method is the analytic process by which concepts are identified and developed in terms of their properties and dimensions. The basic analytic procedures by which this is accomplished are: the asking of questions about the data; and making of comparisons for similarities and differences between each event and other instances of phenomena. Similar events and incidents are labelled and grouped to form categories (Strauss & Corbin, 1990:74). Coding may be applied to a text line by line (See Charmaz, 1995:39), sentence by sentence or paragraph by paragraph, or a code may be linked to the whole text (Flick, 2002:178). The advantage of line by line coding is that it “helps you to refrain from inputting your motives, fears or unresolved personal issues to your respondents and to your collected data” (Charmaz, 1995:37). The approach that a researcher uses will be determined by the content and kind of text being coded. Flick (2002:180) and Bohm (2004:271) provide a useful list of questions that may be asked about the data when coding: What is the concern here? Which phenomenon is mentioned? Which actors are involved? What roles do they play? How do they interact? Which aspects of the phenomenon are mentioned or not addressed? When? How long? Where? (i.e. time, course and location). How intense or strong? What reasons are given or can be reconstructed? With what intentions, to which purpose? What means, tactics and strategies can be used to achieve the goal? Charmaz (2003) also suggests asking more or less the same questions when coding data at this stage. All these questions seem to be based on the set of questions of the data formulated by Glaser (1998:140). Glaser (2004) suggests that researchers should not proceed to selective coding (intermediate in the traditional phase, see Table 8.2) before a potential core category has emerged through theoretical sampling. A core category or substantive code relates to many other categories and their properties, and can explain their variation in a pattern of behaviour (Holton, 2007). b) Intermediate coding (interpretive codes) Open coding may result in hundreds of codes (Strauss & Corbin, 1995:65), depending on the data being analysed. Intermediate coding is applied to condense and discern the categories of codes through constant comparison. As illustrated in Table 8.2 above, intermediate coding is also referred to as selective, axial and focused coding (Mills et al., 2014). The concern at this stage is to establish linkages and connections between categories, in order to understand the phenomenon to which they relate and determine if the data supports emerging categories (Holton, 12 2007; Curtis & Curtis, 2011). This process progresses until saturation point has been reached. The constant comparison process feeds into theoretical sampling, a process that enables the researcher to make a decision regarding what data to collect next, in order to substantiate the emerging theory (Holton, 2007). Diagrams, including matrices, tables, concept maps and cross tabulations, are visual tools that may be utilised in mapping the relationship between categories. c) Advanced coding (theoretical codes) Advanced coding is variously known as theoretical coding or selective coding, as shown in Table 2 above. Novice researchers experience a lot of problems when conducting the theoretical coding process (Holton, 2007; Hernandez, 2009). The core code is selected and identified from all those identified in the first two stages of data analysis. The formulation of a theory and theoretical integration happens during this final phase. All substantive codes/categories are related to the core category by the theoretical code. Theoretical sensitivity, the ability to generate concepts from the data, and relating them to theory assists in conceptual integration (Glaser, 1978:117). Theoretical sensitivity may be improved partly through wide reading (Glaser, 1998:164-165). The storyline is a tool that is used for theoretical integration (Mills et al., 2014). The explanatory power of the storyline is enhanced through the use of theoretical codes, which “are advanced abstractions that provide a framework for enhancing the explanatory power of the storyline and its potential as theory” (Birks & Mills, 2011:123). Predictive statements about the phenomenon under study may be created at this stage if researchers wish to make “predictive statements” – but they may not wish to do so, as they may subscribe to an ontology which suggests that social life is not predictable. 8.5.4 Structured data analysis Structured data analysis procedures are related to qualitative methods such as critical ethnography and hermeneutics. For instance, structured data analysis methods are employed in content analysis, vignettes, Q-methodology and protocol analysis. The data analysis method mainly transforms qualitative data into numbers based on a coding scheme. 8.5.4.1 Content analysis Created by journalists and then adopted by social scientists, content analysis is a research technique that collects and analyses data from texts and messages that are communicated in various ways, including books, newspapers and other physical media (Curtis & Curtis, 2011). Stated differently: Content analysis is a systematic coding and categorising approach you can use to explore large amounts of existing textual information in order to explore large amounts of existing textual information in order to ascertain the trends and patterns of words used, their frequency, their relationships and structures, contexts and discourses of communication (Grbich, 2013:190). Content analysis or content research is “conceptually and logically straightforward” (Curtis & Curtis, 2011:215). Content analysis may either be characterised as enumerative content analysis or ethnographic content analysis. Ethnographic content analysis is concerned with analysing documents for significance and 13 meaning, whereas in enumerative content analysis, the major concern is the frequency of words and categories, including concordance and co-occurrence (Grbich, 2013; Liamputtong, 2013). Qualitative data analysis (QDA) programs such as MAXQDA, NVivo and WordStat 6.1 may be used to conduct content analysis. Patterns are identified and interpreted in doing content research or content analysis. Enumerative content analysis, or what Liamputtong (2013:246) refers to as traditional quantitative content analysis, mainly transforms qualitative data into quantitative forms. The technique tends to overlook the latent or covert elements of messages or text, as its focus is mainly on the obvious elements that can be counted (Curtis & Curtis, 2011). The technique assists in only describing social relations, but cannot explain them. McNabb (2002:414) describes the major advantage and disadvantage of content analysis as the following: it provides the researcher with a structured method for quantifying the contents of a qualitative or interpretive text, and does so in a simple, clear, and easily repeatable format. Its main disadvantage is that it contains a built-in bias of isolating bits of information from their context. Thus, the contextual meaning is often lost or, at the least, made problematic. Conducting content analysis should be partly based on considerations suggested by Grbich (2013:190). Thus, researchers should ensure that they: Have a sufficient number of documents and determine the aspects of the documents to be analysed; Establish the sampling approach when selecting documents; Decide on the level of analysis to be done; Decide on how the codes will be generated; Consider the relationships between concepts, codes and contexts; Record the number of times categories appear; and Ascertain the reliability of the coding scheme. The decoder focuses on the predetermined themes or propositions that are presumed to be in the data or messages. Ethnographic content analysis (ECA) may also start with predetermined categories and themes, like enumerative content analysis, but some categories may emerge during inductive analysis, as Bryman (2012:559) explains: “there is greater potential for refinement of those categories and the generation of new ones”. 8.5.5 Discursive data analysis methods Discursive data analysis procedures are related to qualitative methods such as critical ethnography, historical research and hermeneutics. Flick (2002) also refers to discursive data analysis methods as sequential data analysis procedures. Discursive methods are used to analyse texts, for instance in discourse analysis and semiotic analysis. 8.5.5.1 Discourse analysis Discourse analysis is based on social constructivism assumptions. The fundamental question is framed around how social reality can be understood and explained by investigating discourses about certain situations and processes. LeGreco (2014:69) 14 provides an instructive list of handbooks and introductory texts for discourse analysis. The ability of discourse analysis to deal with language, dialogue and texts has persuaded scholars in fields such as anthropology, communication, sociology, psychology, media studies, rhetoric, education, linguistics and health sciences to explore its role in framing social patterns and practices (Liamputtong, 2013). Researchers doing discourse analysis mainly deal with talk and text. Discourse analysis is close to conversation analysis, another method of studying verbal and non-verbal interaction in context. There are more than fifty ways of doing discourse analysis (Liamputtong, 2013). However, the two dominant discourse analysis approaches are Foucauldian discourse analysis or post-structuralist discourse analysis (Braun & Clarke, 2013), based on Michel Foucault, and critical discourse analysis, which was developed by Norman Fairclough (Grbich, 2013). According to Foucault (1972:49), discourses are not about objects; they do not identify objects, they constitute them and in the practice of doing so conceal their own invention. Power is fundamental to the creation and sustenance of knowledge in society, and any discourse should be understood in the context of power relations. On the other hand, critical discourse analysis is based on the interaction of the text, discursive practices and the social context (Fairclough, 2000). Eriksson and Kovalainen (2008) and Braun and Clarke (2013) write of a third variety of discourse analysis which is prevalent, especially in business and psychology research. The third version of discourse analysis, which draws on constructionist psychology and social psychology, is called social psychological discourse analysis. The aim is to show “how social interaction is performative and persuasive”, and it is a “negotiation about how we should understand the world and ourselves” (Eriksson & Kovalainen, 2008:232). There are many techniques of transcription in discourse analysis, but the Jeffersonian system is gaining popularity (LeGreco, 2014:74). The analysis of text is concerned with dialogue structures, discursive practices and conversation strategies in social settings (LeGreco, 2014:74). Discourse tracing is an important technique for data analysis (LeGreco & Tracy, 2009). Although the procedures for conducting discourse analysis are greatly contested, the steps for using discourse analysis, which are provided by Gill (2000), are instructive (see figure 8.4). 15 Figure 8.4: Steps for using discourse analysis (Adapted from Gill, 2000:178-179) 8.5.6 Instrumental data analysis methods Instrumental methods employ a variety of methods to fulfil “an overarching or ethical commitment” (Madill & Gough, 2008:259). For instance, ethnography may utilise discourse theory and forms of thematic analysis, even if it is committed to naturalistic inquiry. Other examples of qualitative methods that employ instrumental procedures of data analysis include ethnomethodology, feminist research, visual methodologies, action research and media framing analysis (Madill & Gough, 2008). 8.5.6.1 Action research Action research is different from other qualitative research designs because it is action-oriented, solves current problems and empowers the research participants, while extending the frontiers of knowledge. It moves the participants from the realm of being mere subjects to a sphere where they are empowered to understand and positively change their situation. It is an iterative research design proceeding through stages of planning, action and review (Dick, 2014: 51). Action research is prevalent in education, organisational change, community change and farmer research (Dick, 2014:52). Data analysis is done collectively and collaboratively with research participants. Some researchers suggest that even when one is not doing action research as such, it is worth checking one’s draft analyses and interpretations with participants (Chilisa, 2012). Undertaking member checking with the participants to determine whether the themes, arguments or assertions developed from the codes accurately reflect their sentiments enhances descriptive validity (Maxwell, 2005). However, member checking in action research mainly leads to an understanding of a certain situation, and informs action. Some researchers use grounded theory and constant comparison analysis for data analysis in action research (Dick, 2014:52). 16 8.6 Computer-based analysis Qualitative data analysis (QDA) or computer-aided qualitative data analysis software (CAQDAS) programs are increasingly used in the analysis of qualitative data, and they have improved the process of qualitative data analysis. However, computerbased qualitative data analysis software programs have not yet gained full acceptance (Cambra-Fierro & Wilson, 2010:17), despite their potential. Computers may be used in activities such as making field notes, writing up or transcribing notes, editing, sorting, coding, data linking, memoing, storing, searching, indexing and retrieving qualitative materials. There are hopes, fears and fantasies associated with these technologies (Flick, 2002:250; Curtis & Curtis, 2011:51). Some fear that technology might distort qualitative research practice. This fear is unfounded and Flick (2002) regards it as phantasm because QDA software does not conduct the analysis, as the researcher still does the coding. In fact, researchers should be cognisant of the fact that “nothing takes the place of the researcher’s inductive analysis of the raw data” (McMillan & Schumacher, 2014:395). In fact, “creating a coding framework and making decisions about the role of coding in a project still necessitates a great deal of conversation and debates” (Adair & Pastori, 2011:32). Current two-file system theory generating computer programs developed from code and retrieve QDA programs which had single file systems that managed to stored and retrieve data (Grbich, 2013). There are at least 30 different software choices (McMillan & Schumacher, 2014:409). The most popular products for social and management science researchers are Nvivo, previously NUD*IST, and Atlas.ti (Myers, 2009). ATLAS.ti software uses the grounded theory and theoretical coding approaches in modelling the data (Flick, 2002). It can interface with statistical packages such as SPSS©. It has the same capabilities as Nvivo. Paulus, Lester and Dempster (2014) provide a useful list of questions that may guide a researcher when selecting a CAQDAS system: What features will support my analytical approach? Does the system allow me to annotate, link, search, code and visualise data? How does the software assist in data management? What are the benefits and constraints of the software package? Grbich (2013:285) suggests that MAXQDA is a better program than the two in terms of “capability, stability, ease of learning and use”. However, it is important to use a product that is supported by the researcher’s institution, in order to take care of the licensing fees, which at times may be high. Equally important is to determine for the software program was specifically developed. For instance, programs developed from an ethnographic or grounded theory perspective may not be suitable for analysing qualitative data from other contexts. Myers (2009) advises against the use of QDA for hermeneutics and narrative analysis, partly due to the same reasons. 8.7 Interpretation of qualitative data: illustrating assertions and interpretations through results Interpretation is the terminal phase of qualitative inquiry (Denzin & Lincoln, 2005:909). Discussing the interpretation of data in qualitative researchis difficult, as it 17 is a hotly contested area. The major challenge in discussing the interpretation of qualitative data stems from the fact that interpretation is regarded as an art that is not amenable to formal rules, as the “processes that define the practices of interpretation and representation are always ongoing, emergent, unpredictable, and unfinished” (Denzin & Lincoln, 2005:909). However, with reference to Liamputtong and Ezzy (2005), we have attempted to formalise the notion of interpretation. The interpretation of data is the core of qualitative research (Flick, 2002:176). This phase entails the assessment, analysis and interpretation of the empirical evidence that has been collected. The different points of view of the participants are presented in sufficient detail and depth, so that the reader may be able to gauge the accuracy of the analysis. Stated differently, a thick description is presented in the form of an “analytical narrative” (McMillan & Schumacher, 2014:361). The data are used to illustrate and validate the interpretation of the data. Pertinent words and comments of the participants are usually quoted. Verbatim quotations of the participants assist in “revealing how meanings are expressed in the respondents’ words rather than the words of the researcher” (Baxter & Eyles, 1997:508). Chenail (2012:1) cautions that qualitative researchers should be able to refer to their original data and be able “to construct evidence of the code from the data”. Furthermore, qualitative researchers should neither say more than what the data says nor less than the data before them, as they: need to be aware of making errors of deficiency and exuberance in reporting our qualitative analysis of the quality we create from the data. By deficiency I mean “Don’t try to say less than what the data show” and by exuberance I mean, “Don’t try to say more than what data show.” (Chenail, 2012:1). The researcher’s perceptions, biases and personal beliefs should also be accounted for. In other words, the interpretation “includes the voices of participants, the reflectivity of the researcher, and a complex description ... of the problem” (Creswell, 2007:37). Points related to previous research are good candidates for the interpretation section of a qualitative report. Although qualitative presentations are mainly in narrative form, statistical tools such as descriptive statistics may be used to summarise data. The conclusions should be consistent with the findings. The synthesis of the findings may be followed by the suggestion of a model or theory. Models and theories were the subject of Chapter Four. 8.8 Criteria for evaluating qualitative research Some scholars question the value of qualitative research (see Biesta, 2007; Hammersley, 2007). It is noteworthy that: much of the pressure for qualitative criteria comes not so much from the context of researchers judging research, or even students learning to do this, but rather from that of lay ‘users’ of research (notably policymakers and practitioners) assessing its quality (Hammersley, 2007:289). There is a perception that qualitative research is not rigorous, as the claim is that its methods and processes are not rigidly controlled, unlike quantitative research, which is subject to strict rules and standards in relation to the methodology that is used (Liamputtong, 2013; Birks, 2014; Hammersley, 2007). Rigour is fundamental to any 18 research enterprise because it addresses matters of the quality of the research, including the analysis and interpretation of generated data. Tobin and Begley (2004:390) write that: “Without rigour, there is as a danger that research may become fictional, worthless, as contributing to knowledge”. Although qualitative research is characterised by methodological pluralism, as alluded to in section 8.3, Hammersley (2008), Spencer et al. (2003) and Tracy (2010) attempted to identify criteria by which qualitative research may be assessed. There is no agreement on whether or not a single set of criteria for assessing rigour in qualitative research is possible, as a result of the multitude of research designs used in qualitative research. However, Hammersley (2007:300) advises that guidelines for valuating qualitative research are useful, as long as they do not become “a substitute for the practical capacity to assess research”. On the other hand, Tracy (2010:837) states that any model for gauging quality in qualitative research should “leave space for dialogue, imagination, growth and improvisation”. With reference to Denzin and Lincoln (2005), the criteria for evaluating qualitative research by looking at the results in relationship to the foundations of truth and knowledge, or the nature of reality (i.e. epistemologies and ontological realism, respectively) are suggested. The knowledge claims criteria are based on what Denzin and Lincoln (2005:909) term foundationalists, quasi-foundationalists and nonfoundationalists positions. Many researchers who are interested in understanding the criteria for evaluating qualitative research have found the categories to be useful in this regard. Foundationalists are of the opinion that the same criteria that are applied to quantitative research should be used to evaluate qualitative inquiry. In evaluating qualitative research, foundational scholars use the variants of the classical criteria, such as internal validity, external validity, reliability and objectivity, which are embedded in the positivist and post-positivist paradigms (Denzin & Lincoln, 2005; Hammersley, 2007). Quasi-foundationalists maintain that the evaluative criteria should be unique and rooted in the constructivist epistemology. Consequently, qualitative research should be evaluated in terms of plausibility (i.e. is the claim plausible?), credibility (i.e. is the claim informed by credible evidence?), and relevance (i.e. what is the claim’s relevance to knowledge about the world?) (Hammersley, 1995). This typology underscores the need for a study to demonstrate trustworthiness, as articulated by Lincoln and Guba (1985). Maintaining the audit trail of the research project may enhance the credibility, plausibility, authenticity and dependability of a qualitative study. Although Cutcliffe and McKenna (2004) argue to the contrary and elevate the expertise of the researcher as a major determinant of credibility and authenticity, Birks (2014) strongly argues that expertise does not negate “the need to demonstrate precision in research work”. Maintaining an audit trail fosters procedural precision and transparent accountability. It also enhances procedural reliability (Flick, 2002; Liamputtong, 2013) and ensures the dependability of the data and the analysis thereof. By recording the research activities, any changes to the research plans and the reasons for any deviations and exceptions do not only demonstrate professionalism, but make the readers have confidence in your results (Birks, 2014) 19 and adjudge them to be credible. Documenting data collection and analysis processes ensures that qualitative research is not “limited to a mechanistic analysis and reporting of content” (Cambra-Fierro & Wilson, 2010:18). Documenting all the steps and decisions taken during data collection and analysis provides the researcher with an opportunity to deal with and report on the data that “jump out” in contradiction. Non-foundationalists arise as a result of the influence of the feminist and communitarian ethic of empowerment, community and moral solidarity. They contend that the empirical claim to knowledge cannot be done epistemologically because social science research serves a moral and political purpose in a given context (Denzin & Lincoln, 2005:911). In other words, valid research should address matters of social inequality and improve the lives of the marginalised. A framework for evaluating qualitative research should provide some value-relevant criteria, which can be used to judge the validity and trustworthiness of any qualitative research. Of course, this constitutes a tacit critique of Hammersley (2007), who does not wish to see the introduction of value criteria as part of the process of doing research. 8.9 Conclusion We have demonstrated in this chapter that qualitative data analysis involves the identification, examination, comparison and interpretation of patterns and themes. We have shown that qualitative data analysis is different to quantitative traditions of analysis, hence requiring a different approach to the one discussed in Chapter Seven when making sense of the data. We established that there are as many sources of qualitative data as there are types of approaches used to analyse qualitative data. We have described a typology of qualitative data analysis which includes discursive, thematic, structured and instrumental methods of data analysis. Computer-based analysis of qualitative data was also explained. We then reflected on the interpretation of qualitative data and concluded by discussing the criteria for evaluating qualitative research. Window into understanding qualitative data analysis methods Activity 8.1 Take a few minutes to jot down a few words that describe your understanding of the various qualitative data analysis procedures outlined in this chapter. Activity 8.2 On the one hand, the commonly used qualitative methods used in business and management are action research, case study, ethnography and grounded theory (Myers, 2009:29). On the other hand, content analysis, grounded theory and discourse analysis are establishing themselves in the field of psychology, as demonstrated by Carrera-Fernández, Guàrdia-Olmos and Peró-Cebollero (2014). Trace the prevalent qualitative methods of data analysis in your field. Activity 8.3 How do researchers achieve rigour in qualitative research? Activity 8.4 To what extent can computer-based analysis eventually replace researchers in the inductive analysis of qualitative data? References Adair, J.K. & Pastori, G. (2011). Developing qualitative coding frameworks for educational research: immigration, education and the Children Crossing Borders project. International Journal of Research and Methods in Education, 34(1): 31-47. 20 View publication stats