Analytical Procedures in Qualitative Research Grounded Theory & Phenomenological Research

Ground Theory and Phenomenological  Analytics

Analytical procedures in Qualitative Research  Analytic Procedures Grounded Theory Analysis Glaser and Strauss' Grounded Theory Method  Phenomenological Analysis 

Analytic Procedures

    Data management in qualitative research is reductive in nature because it involves converting large masses of data into smaller, more manageable segments. By contrast, qualitative data analysis is constructionist: It involves putting segments together into a meaningful conceptual pattern. Qualitative analysis is an inductive process that involves determining the pervasiveness of key ideas. 

    Although there are various approaches to qualitative data analysis, some elements are common to several of them. We provide some general guidelines, followed by a description of the procedures used by grounded theory researchers, phenomenologists, and ethnographic researchers. We also provide information about analyzing data from focus group interviews and briefly note a strategy for analyzing triangulated qualitative and quantitative active data

A General Analytical Overview

    The analysis of qualitative materials usually begins with a search for themes. DeSantis and Ugarriza (2000), in their thorough review of the way which the term theme is used among qualitative researchers, offer this definition of theme: “A theme is an abstract entity that brings meaning and identity to a current experience and its variant manifestations. As such, a theme captures and unifies the nature or basis of the experience into a meaningful whole”. 

    Themes emerge from the data. Themes may develop within categories of data (ie, within categories of the coding scheme used for indexing materials), but may also cut across them. For example, in Polit and colleagues' (2000) study, one theme was that these single mothers took great pride in their own resourcefulness in accessing food services (A codes) and developing strategies to avoid hunger (C codes) for their families. 

    The search for themes involves not only discovering commonalities across participants, but also seeking natural variation. Themes are never universal. Researchers must attend not only to what themes arise but also to how they are patterned. Does the theme apply only to certain types of people or in certain communities? In certain contexts? At certain periods? What are the conditions that precede the observed phenomenon, and what are the apparent consequences of it? 

    In other words, the qualitative analyst must be sensitive to relationship within the data. Researchers' search for themes, regularities, and patterns in the data can sometimes be facilitated by charting devices that enable them to summarize the evolution of behaviors, events, and processes. For example, for qualitative studies that focus on dynamic experiences such as decision-making it is often useful to develop flow charts or time-lines that highlight time sequences, major decision points and events, and factors affecting the decisions.

    In this phase, the concern is whether the themes inferred accurately represent the perspectives of the people interviewed or observed. Several validation procedures can be used, as discussed . If more than one researcher is working on the study, sessions in which the themes are reviewed and specific cases discussed can be highly productive. Investigator triangulation cannot ensure the matric validity, but it can minimize idiosyncratic biases. 

    Using an iterative approach is almost always necessary. That is, researchers derive themes from the narrative materials, go back to the materials with the themes in mind to see if the materials really do fit, and then refine the themes as necessary. It is also useful to undertake member checks that is, to present the preliminary thematic analysis to some study participants, who can be encouraged to offer comments to support, contradict, or modify the thematic analysis. 

    At this point some researchers introduce quasistatistics a tabulation of the frequency with which certain themes, relations, or insights are supported by the data. The frequencies cannot be interpreted in the same way as frequencies generated in survey studies, because of imprecision in the sampling of cases and enumeration of the themes. Nevertheless, as Becker (1970) pointed out, quasi-statistics may allow the investigator to dispose of certain troublesome null hypotheses. 

    A simple frequency count of the number of times a given phenomenon appears may make untenable the null hypothesis that the phenomenon is infrequent. A comparison of the number of such instances with the number of negative cases instances in which some alternative phenomenon that would not be predicted by his theory appears may make possible a stronger conclusion, especially if the theory was developed early enough in the observational period to allow a systematic search for negative cases. Sandelowski (2001) believes that numbers are underutilized in qualitative research because of two myths: 

(1) real qualitative researchers do not count, and (2) qualitative researchers cannot count. Numbers are helpful in highlighting the complexity and work of qualitative research and in generating meaning from the data. Numbers are also useful in documenting and testing interpretations and conclusions and in describing events and experiences. Sandelowski warns, however , of the pitfalls of overcounting whole. 

    The various themes need to be interrelated to provide an overall structure (such as a theory or integrated description) to the data. The integration task is a difficult one, because it demands creativity and intellectual rigor if it is to be successful. One strategy that sometimes helps in this task is to cross-tabulate dimensions that have emerged in the thematic analysis.

    In the remainder of this section, we discuss analytic procedures that have been adopted by grounded theory researchers, phenomenologists, and ethnographers. It should be noted, however, that qualitative researchers who conduct studies that are not based in a formal research tradition may simply say that a content analysis was performed. Qualitative content analysis is the analysis of the content of narrative data to identify prominent themes, and patterns among the themes primarily using an analysis style that can be characterized as either template analysis or editing analysis.

Grounded Theory Analysis

    The general analytical procedures just described provide an overview of how qualitative researchers make sense of their data and distill from them insights into processes and behaviors operating in naturalistic settings. 

    However, variations in the goals and philosophies of qualitative researchers also lead to variations in analytic strategies. This section describes data analysis in grounded theory studies. As noted , one grounded theory approach was developed by Strauss and Corbin (1998), and another is Glaser and Strauss' (1967) grounded theory method of generating theories from data.

Glaser and Strauss' Grounded Theory Method

    Grounded theory uses the constant comparative method of data analysis. This method involves a comparison of elements present in one data source (eg, in one interview) with those identified in another. The process is continued until the content of each source has been compared with the content in all sources. In this fashion, commonalities are identified. The concept of fit is an important element in grounded theory analysis. 

    Fit is the process of identifying characteristics of one piece of data and comparing them with the characteristics of another datum to determine if they are similar (Morse & Singleton, 2001). In the analytical process, fit is used to sort and reduce data. Fit enables the researcher to determine if data can be placed in the same category or if they can be related to one another.

     Glaser (1992) warns qualitative researchers not to force an analytic fit when it is not present in the data. He stated that “if you torture data enough it will give up!”. Forcing a fit hinders the development of a relevant theory. Coding in Glaser and Strauss' grounded theory approach is used to conceptualize data into patterns or concepts. The empirical substance of the topic being studied is conceptualized by substantive codes, whereas theoretical codes conceptualize how the substantive codes relate to each other. 

    In the Glaser and Strauss approach, there are two types of substantive codes: open and selective. Open coding, used in the first stage of the constant comparative analysis, captures what is going on in the data. Open codes may be the actual words used by the participants. Through open coding, data are broken down into incidents and their similarities and differences are examined. During open coding, researchers ask “What category or property of a category does this incident indicate?” (Glaser, 1978,). There are three different levels of open coding, depending on the level of abstraction. 

    Level I codes (sometimes called in vivo codes) are derived directly from the language of the substance active area. They have vivid imagery and “grab.” As researchers constantly compare new level I codes with previously identified ones, they condense them into broader categories (level II codes). Theoretical constructs, also called level III codes, are the most abstract level of codes. These con structs “add scope beyond local meanings” (Glaser, 1978,) to the generated theory. Collapsing of level II codes aids in identifying constructs. 

    Open coding ends when the core category is discovered, and then selective coding begins. The core category is a pattern of behavior that is relevant or problematic for study participants. In select active coding, researchers code only those data that are related to the core variable. One kind of core variable can be a basic social process (BSP) that evolves over time in two or more phases. All BSPs are core variables, but not all core variables have to be BSPs. To help researchers decide on a core category, Glaser (1978,) provided nine criteria:

1. It must be central, meaning that it is related to many categories.

2. It must reoccur frequently in the data.

3. It takes more time to saturate than other categories.

4. It relates meaningfully and easily to other categories.

5. It has clear and grabbing implications for formal theory.

6. It has considerable carry-through.

7. It is completely variable.

8. It is a dimension of the problem.

9. It can be any kind of theoretical code. 

    Theoretical codes help the grounded theorist to weave the broken pieces of data back together again. Throughout coding and analysis, grounded theory analysts document their ideas about the data, themes, and emerging conceptual scheme in memos. Memos preserve ideas that may initially not seem productive but may later prove valuable once further developed. Memos also encourage researchers to reflect on and describe patterns in the data, relationships between categories, and emergent conceptualizations.

    Glaser and Strauss' grounded theory method is concerned with the generation of categories, properties, and hypotheses rather than testing them. The product of the typical grounded theory analysis is a conceptual or theoretical model that endeavors to explain a pattern of behavior that is both relevant and problematic for study participants. 

    Once the basic problem emerges, the grounded theorist then goes on to discover the process these participants experience in coping with or resolving this problem in the process of developing grounded theory, the concept of emergent fit can help to prevent individual substantive theories from being “respected little islands of knowledge” (Glaser, 1978, ). 

    As Glaser pointed out, generating grounded theory does not necessarily require discovering all new categories or ignoring ones promiscuously identified in the literature: “The task is, rather, to develop an emergent fit between the data and a pre-existent category that might work. Therefore, as in the refitting of a generated category as data emerge, so must an extant category be carefully fitted as data emerge to be sure it works. 

    In the bargain, like the generated category, it may be modified to fit and work. In this sense the extant category was not merely borrowed but earned its way into the emerging theory” Emergent fit does not imply relevant existing literature should be disregarded. Through constant comparison, one compares concepts emerging from the data with similar concepts from existing theory or research to determine which parts have emergent fit with the theory being generated.

    Strauss and Corbin's approach The Strauss and Corbin (1998) approach to grounded theory analysis differs from the original Glaser and Strauss method with regard to method and outcomes. Glaser (1978) stressed that to generate a grounded theory, the basic problem must arise from the data it must be discovered. The theory is, from the very start, grounded in the data, rather than starting with a preconceived problem. Strauss and Corbin, however, state that the research itself is only one of four possible sources of the research problem. 

    Research problems can, for example, come from the literature or a researcher's personal and professional experience. The Strauss and Corbin method involves three types of coding: open, axial, and selective coding. In open coding, data are broken down into parts and compared for similarities and differences. Similar actions, events, and objects are grouped together as more abstract concepts, which are called categories. 

    In open coding, the researcher focuses on generating categories and their properties and dimensions. In axial coding, the analyst develops categories and links them with subcategories. Strauss and Corbin (1998) term this process of relating categories and their subcategories as “axial because coding occurs around the axis of a category, linking categories at the level of properties and dimensions”. What is called the paradigm is used to help identify linkages among categories. 

    The basic components of the paradigm include conditions, actions/interactions, and consequences. Selective coding is the process in which the findings are integrated and refined. The first step in integrating the findings is to decide on what Strauss and Corbin term the central category (sometimes called the core category ), which is the main theme of the research. 

    Recommended techniques to facilitate identifying the central category are writing the storyline, using diagrams, and reviewing and organizing memos. The outcome of the Strauss and Corbin approach is, as Glaser (1992) terms it, a fully preconceived conceptual description. On the other hand, the original grounded theory method (Glaser & Strauss, 1967) generates a theory that explains how a basic social problem that emerged from the data is processed in a social scene.

 Phenomenological Analysis 

    Schools of phenomenology have developed different approaches to data analysis. Three frequently used methods for descriptive phenomenology are the methods of Colaizzi (1978), Giorgi (1985), and Van Kaam (1966), all of whom are from the Duquesne school of phenomenology, based on Husserl's philosophy .The basic outcome of all three methods is the description of the meaning of an experience, often through the identification of essential themes. 

    Phenomenologists search for common patterns shared by particular instances. There are, however, some important differences among these three approaches. Colaizzi's method, for example, is the only one that calls for a validation of results by returning to study participants. Giorgi's analysis relies solely on researchers. His view is that it is inappropriate either to return to participants to validate findings or to use external judges to review the analysis. Van Kaam's method requires that intersubjective agreement be reached with other expert judges.

    A second school of phenomenology is the Utrecht School. Phenomenologists using this Dutch approach combine characteristics of descriptive and interpretive phenomenology. Van Manen's (1990) method is an example of this combined approach in which researchers try to grasp the essential meaning of the experience being studied. According to Van Manen , thematic aspects of experience can be uncovered or isolated from participants' descriptions of the experience by three methods :

(1) the holistic approach,

(2) the selective or highlighting approach, and

(3) the detailed or lineby-line approach.

    In the holistic approach, researchers view the text as a whole and try to capture its meanings. In the selective approach, researchers highlight or pullout statements or phrases that seem essential to the experience under study. In the detailed approach, researchers analyze every sentence. Once themes have been identified, they become the objects of reflection and interpretation through follow-up interviews with participants. 

    Through this process, essential themes are discovered. In addition to identifying themes from the participants' descriptions, Van Manen also calls for gleaning thematic descriptions from artistic sources. Van Manen urges qualitative researchers to keep in mind that poetry, literature, music, painting, and other art forms can provide a wealth of experiences that can be used to increase insights in the reflection process as the phenomenologist tries to interpret and grasp the essential meaning of the experience being studied. 

    These experiential descriptions in literature and art help challenge and stretch the phenomenologist's interpretive sensibilities.A third school of phenomenology is an interpretive approach called Heideggerian hermeneutics . Diekelmann , Allen, and Tanner (1989) have described a seven-stage process of data analysis, the outcome of which is a description of shared practices and common meanings. Diekelmann and colleagues' stages of data analysis include:

1. All the interviews or texts are read for an overall understanding.

2. Interpretive summaries of each interview are written.

3. A team of researchers analyzes selected transcribed interviews or texts.

4. Any disagreements on interpretation are resolved by going back to the text.

5. Common meanings and shared practices are identified by comparing and contrasting the text.

6. Relationships among themes emerge.

7. A draft of the themes along with examples from texts are presented to the team. Responses or suggestions are incorporated into the final draft.

    Pollio, Henley, and Thompson (1997) propose another method for conducting a hermeneutic phenomenological study. Their method begins with bracketing. Their bracketing is not, however, viewed as a subtractive process of removing one's presuppositions, but instead as a positive process, a way of seeing. 

    Instead of suspending preconceived notions, as described by Husserl, Pollio, and colleagues call for researchers to apply a world view. Pollio and colleagues' method begins with a bracketing interview. The researcher is the first person to be interviewed about the topic under study which raises his or her awareness of presupposition. Once interviews have been conducted and transcribed, the hermeneutic circle begins. 

    This is an interpretive process of continuously relating a part of the text (the transcribed interview) to the whole of the text. Pollio and colleagues described three types of interpretation: group, idiographic (particular), and nomothetic (general). In group interpretation, a transcript is read aloud. 

    Meanings and relationships among meanings are discussed. After one transcript is interpreted, the remaining transcripts are usually interpreted by the primary researcher. At certain times the researcher goes back to the group with idiographic descriptions and nomothetic themes. The group provides feedback on whether the descriptions and themes are supported by the data. Each transcript is interpreted in the context of all other interview transcripts.