1 1 Using NVivo to Analyze Qualitative Data Anette M. DeNardo, Ed.D. California University of Pennsylvania Lisa Lopez Levers, Ph.D. Duquesne University Abstract Analyzing data gathered by qualitative means audio taped interviews, video taped focus groups, researcher field notes, and others can be an overwhelming task. There are no established formulae for transforming the data into findings. The challenge of the process is to make sense of massive quantities of data. The process involves sifting through the data, filtering out the significant information, identifying patterns, and constructing a framework for communicating the essence of what is revealed. That process can be assisted by the use of computer software to facilitate the storage, coding, retrieval, comparison, and linking of that data. Software can ease the laborious task which would otherwise be performed manually. NVivo by Qualitative Solutions and Research Pty. Ltd. of Melbourne, Australia is one of the software packages designed for this purpose. The presenters discussed the use of NVivo in analyzing qualitative data collected in a research study. Using NVivo to Analyze Qualitative Data Data analysis is probably that part of qualitative research that most clearly differentiates it from quantitative research methods. It is also the least understood aspect of qualitative research, particularly for researchers familiar with traditional quantitative methods (Maxwell, 1996). While the goal of qualitative analysis is the transformation of data into findings, no
2 2 formulae to guide that transformation exist (Patton, 2002). The challenge lies in making sense of huge amounts of data by reducing the volume of raw information, sifting trivia from significance, identifying significant patterns, and constructing a framework for communicating the essence of what the data reveal (Patton, p. 432). One way to arrive at the findings is to identify patterns in the data. However, finding uncertainties and ambiguities is another. A related issue is deciding when to commence the analysis of qualitative data. Maxwell (1996) noted that a common problem in qualitative studies is permitting unanalyzed data to accumulate, causing the final analysis to be a much more difficult and disheartening task. He went on to state that data analysis should begin immediately after the first interview or observation terminates and continue from there until the analysis is complete. This sentiment was echoed by Merriam (2002). The amount of data generated by qualitative methods is extremely large, particularly when compared with traditional quantitative data collections, and making sense of pages and pages of interviews and field notes can be overwhelming. Organizing and analyzing the data can appear to be an impossible task (Patton, 2002). One way to accomplish this daunting task is by hand, a method still employed by some qualitative researchers. However, more than a few noted qualitative theorists have encouraged the use of qualitative data analysis software tools designed to manage data more efficiently throughout the course of a research project (e.g., Berg, 2001; Denzin & Lincoln, 1998; Kelle, 1997a, 1997b; Krueger, 1998; Merriam, 2001; Miles & Hueberman, 1994; Morse & Richards, 2002; Patton, 2002; Silverman, 2000, 2001; Taylor & Bogdan, 1998; Tesch, 1990). Whether one performs a traditional, hands-on analysis of data or uses computer software to assist in the process, certain steps must be taken. First, the data are collected and transformed into text (Berg, 2001). This first step must be undertaken by humans. To date, voice recognition software is not up to the task of translating the statements of multiple voices into text. Next, codes are developed or identified in the data and affixed to the textually-represented data (Berg). This, again, is a human task whether it is accomplished by hand or through the use of computer
3 3 software. Third, the codes are translated into categorical themes. Fourth, the collected materials are sorted using these categories. Fifth, those same materials are closely examined in order to isolate meaningful patterns (Berg). These last three tasks can be greatly eased by the use of software. Lastly, those identified patterns are reviewed in light of previous research and existing theories, yielding a small set of generalizations (Berg). Here, again, the human must intervene. It cannot be emphasized enough that although software programs can be used to facilitate data storage, coding, retrieval, comparing, and linking, only humans can perform the difficult task of analysis (Fielding, 1994; Patton, 2002). This difficulty is caused, in large part, by the richness of qualitative data (Fielding). Computers and software are simply tools that can be used to assist in the process. They have the capacity to significantly ease the laborious task of manually performing the analysis. They can speed up the process of locating coded themes, grouping data together in categories, and comparing passages in transcripts or incidents from field notes (Patton, p. 442). Yet, the researcher must ultimately decide which themes have emerged, what name should be attached to each theme, and the meanings that are extracted. Software that assists in the analysis of qualitative data is referred to as Computer-Assisted Qualitative Data Analysis Software (CAQDAS) (Patton, 2002). There are three basic types of software used in the analysis of qualitative data. Some only retrieve text; others code and retrieve text; still others are referred to as theory-building software (Fielding, 1994). Text retrieving software packages permit the user to recover data based on keywords that appear in the text. They locate these keywords, even when they are in combination with other words, and can search one or more files. Many times these packages can mark the text, separate the marked text into a new file, count the number of occurrences of the keywords, display the keywords in their context, and organize the retrieved pieces of text in some format. These software packages are very fast and efficient (Fielding, 1994). Code-and-retrieve software packages help the researcher separate text into segments, attach codes to those segments of text, and then find and display all text segments with a given
4 4 code or some combination of them. This process reflects the hands-on method used by qualitative researchers in which paper, scissors, markers, and tape or glue was used to mark text, cut apart sections of it, and, after much sorting and reorganizing, put back together those sections of text that related to a particular theme. Even the weakest of these types of software packages are more systematic and thorough than the manual equivalent (Fielding, 1994). Most of these types of packages have been developed by qualitative researchers and are based in grounded theory (Fielding); that is, theory that is inductively generated from fieldwork,... that emerges from the researcher s observations and interviews out in the real world rather than in the laboratory or the academy (Patton, 2002, p. 11). Because of this, however, they are often lacking in support, documentation, and availability (Fielding). Theory-building software is concerned with more than building categories; it deals with the relationships between the categories themselves. Using this type of software, the researcher can make connections between nodes, develop higher-order categories and classifications, formulate assertions that fit the data, or test those assertions to determine if they apply (Fielding, 1994). This is accomplished via the use of information retrieval functions (Kelle, 1997a). Theory-building software is often based on formal logic or organized around a set of rules and offers Boolean searching and hypothesis-testing features (Fielding). Coding is essential in qualitative analysis (Fielding, 1994; Kelle, 1997a; Patton, 2002). This process usually begins when the researcher identifies major themes and the sections of text in which those themes reside. Each of the identified regions is marked with a relevant code (Kelle). Although software packages provide a variety of tools and formats for coding, the principle is the same as when it is done manually (Patton). In the software, these codes are stored along with the location, or address, of the appropriate passage of text so that the researcher can locate all the information associated with a certain topic. When hypotheses are tested, the software package allows the researcher to enter a Boolean-type hypothesis in the form of an if-then statement. For example, IF code-a AND (code-b OR code-c) AND NOT code-d could be used to determine the inclusion and exclusion of certain codes. Based on the result, the
5 5 researcher can conclude that a certain hypothesis must be accepted (Kelle) or rejected. In general, while these programs mark text, build codes, index, categorize, create memos, and display multiple text entries side-by-side (Patton), they offer the researcher the chance to play with the data and possibly open up new perspectives and stimulate new insights (Kelle). These packages also differ in several crucial ways. How the data is entered, where it is stored, how codes are assigned, how codes are organized, whether or not memos can be attached to codes, how data linking occurs, how easy the data is navigated and browsed, and the ease and speed with which searching and retrieval occurs are some of the areas in which differences occur (Fielding, 1995). Some packages are designed for one user and others for multiple users with varying degrees of access to the data (Patton, 2002). As Miles and Hueberman (1994) point out, it is not a matter of which computer software program is the best, but rather, it is a matter of the researcher s level of comfort using a computer and the particular purpose for using the specific software program (Taylor & Bogdan, 1998). The use of computer software in qualitative research is becoming more popular because the computer has the capacity for organizing massive amounts of data, as well as facilitating communication among members of a research team (Merriam, 2002, p. 166). But, before the data can be analyzed, it must be prepared. This process involves typing notes, transcribing interviews, and entering other data from which the researcher will be working (Merriam). Often, a standard word processor is the best tool to use in creating clean records from which to work. Usually, the data prepared in this fashion can be used in conjunction with another program to assist in the analysis phase. Once the data has been entered, it must be divided into meaningful segments that are easy to locate. Here, programs specifically developed with qualitative research in mind are most helpful. Using these types of programs, the researcher can search for, sort, retrieve, and rearrange data segments (Merriam, 2002). Most software packages were designed to perform specific qualitative tasks; herein lie their strengths. Unfortunately, these same products suffer from usability issues common to products with limited audiences, they often lack compatibility
6 6 with other commercial programs, and they often have steep learning curves. In the last 10 years, the number of software programs specifically designed for qualitative research has increased. The difficulty for most researchers is that of selecting the most appropriate software (Merriam). The use of computer software in qualitative research has many advantages. Searches can be done with great speed, and they are more comprehensive than those that can be completed by hand. These software packages can cope with overlapping codes and multiple codes. They can conduct multiple searches and search for more than one code at the same time. They can also be used to attach memos at certain points of the text (Coffey, Holbrook, & Atkinson, 1996). Yet, there may be problems associated with the use of software in qualitative research. In the future, qualitative researchers will probably analyze larger documents and a greater number of them (Plass & Schetsche, 2000). This will probably occur to increase the reliability of the findings of qualitative research and because of the proliferation of software to assist in the process (Coffey et al., 1996). Coffey et al. fear that this increased use of software in the analysis of qualitative research could lead to a new orientation in the qualitative research field that of homogeneity. That is, researchers may adopt a particular set of strategies used in the analysis of qualitative data, which may take away from the traditional, open-ended analysis approach which has, to this point in time, been used by qualitative researchers in analyzing their data. Kelle (1997b), however, contends that the computer has been and will continue to be used to help in manipulating large databases of information in order to assist the qualitative researcher in performing the analysis of the data and does not foresee the same dilemma as Coffey et al. One of the popular packages used for qualitative research purposes is NVivo by Qualitative Solutions and Research Pty. Ltd. of Melbourne, Australia. A recent text (Morse & Richards, 2002) has offered the beginning researcher an overview of data creating decisions and explanations for which analytic tools fit best. The book comes with a software demo so that the reader can sample some of the tools available in QSR s NVivo. NVivo combines the coding of rich data with familiar ways of editing and revising rich text (Richards, 1999, p. 1) and is easier
7 7 to introduce to students than other packages. Also, the software can be learned during the actual research process rather than in lengthy preparatory training phases. NVivo provides a variety of tools for manipulating data records, browsing them, coding them, and annotating and gaining access to data records quickly and accurately (Richards, 1999). NVivo has tools for recording and linking ideas in many ways, and for searching and exploring the patterns of data and ideas. It is designed to remove rigid divisions between data and interpretation, should that be the researcher s goal. It offers many ways of connecting the parts of a project, integrating reflection and recorded data (Richards, p. 4). As the researcher links, codes, shapes and models data, NVivo assists in the management and synthesis of ideas. It offers a range of tools for pursuing new understandings and theories about the data and for the construction and testing of answers to research questions. In order to accomplish the research tasks necessary to use NVivo, the researcher must create a project. The project will contain the data, ideas about it, and the links between the various components. A project can be as simple or complex as the researcher wishes. Within a project, data can be managed within three systems documents, nodes, and attributes. The various actions that can be performed bring these three systems together and, in doing so, give values to attributes, linking, coding, and shaping in sets the documents and nodes. The documents, nodes, or attribute value can be searched with the user specifying the scope of the search. This work is begun from what is called the Launch Pad. It is a pop-up window that serves as NVivo s control center and provides easy access to the user in creating projects and opening them (Richards, 1999). Once a project is opened or created, the user is presented with another window, the Project Pad. This Project Pad contains tabs and buttons that represent the structure of a project and give the user entree to the most common activities associated with the package (Richards, 2002). Documents are rich text records created within or imported into the package. Using the Project Pad, documents can be explored, browsed, changed, linked, and coded. Visual coding
8 8 can be used to mark passages for review. For example, everything on a certain topic can be marked in the same color. Nodes are containers for categories and coding. They can represent abstract ideas, concepts, people, places, processes, or any other category established within a project. Nodes can contain any amount of document coding. Once in the Node Explorer, the user can drag and drop nodes in order to reorganize them and change the index system as ideas form and merge. Some nodes are created by the researcher; others are created automatically by the program (Richards, 2002). Attributes store information and objects about data sources, an organization, the people, a site, or other aspects of a qualitative research study. Examples include the year a document was created, the gender of a respondent, the score of a participant on a certain scale, the name of the interviewer, the location at which a photograph was taken. An attribute can be used to store information about the documents collected and created or to store information about nodes. When searching occurs, attributes are integrated so that the information stored there is accessible in the search. Attributes can be numbers, character strings, Boolean values, or dates. The user is able to specify the range of values for attributes (Richards, 2002). Normally, qualitative researchers link data and ideas in order to express their understanding of certain cultures. There are three ways in which NVivo permits such linking DataBites, DocLinks, and NodeLinks. DataBites are links made at selected text within a document. The user selects the text and easily links that text to a photograph, document, etc. DocLinks and NodeLinks can be placed in a document, anywhere within the text, or at a node. A DocLink can take the user to an existing project document or documents or to a new document. A NodeLink can take the user to text coded at an existing node or nodes or to textual passages. By using these linking techniques, the user can make internal annotations or write memos concerning certain pieces of data (Richards, 2002). Coding is accomplished in NVivo by placing references to text at nodes. NVivo supports several modes of coding, all of which are fast and efficient. Visual coding can be accomplished
9 9 by changing the font, color, and appearance of text. In other forms of coding, the user selects specific text and uses the appropriate button or performs an indicated action (Richards, 2002). A set is a collection of objects. Documents or nodes can be arranged in any number of sets. Once a set is created, the user can code within a set, search within a set, etc. Sets can be named, made, and unmade. The icon for a document or node is simply dragged into the appropriate set (Richards, 2002). This is just a brief introduction of some of the functions that can be performed using NVivo. With practice, qualitative researchers can expedite the analysis of the enormous sets of data they encounter.
10 10 References Berg, B. L. (2001). Qualitative research methods for the social sciences (4th ed.). Boston: Allyn & Bacon. Coffey, A., Holbrook, B., & Atkinson, P. (1996). Qualitative data analysis: Technologies and representations. Sociological Research Online, 1(1). Retrieved from Denzin, N. K., & Lincoln, Y. S. (1998). Strategies of qualitative inquiry. Thousand Oaks, CA: Sage. Fielding, N. G. (1995, May). Choosing the right qualitative software package. Data Archive Bulletin, 58. Retrieved from Fielding, N. G. (1994, September). Getting into computer-aided qualitative data analysis. Data Archive Bulletin. Retrieved from Kelle, U. (1997a, May). Capabilities for theory building and hypothesis testing in software for computer-aided qualitative data analysis. Data Archive Bulletin, 65. Retrieved from [Word Document]. Kelle, U. (1997b). Theory building in qualitative research and computer programs for the management of textual data. Sociological Research Online, 2(2). Retrieved from Krueger, R. A. (1998). Analyzing and reporting focus group results. Thousand Oaks, CA: Sage. Maxwell, J. A. (1996). Qualitative research design: An interactive approach. Thousand Oaks, CA: Sage. Merriam, S. B. (2001). Qualitative research and case study applications in education (2nd ed.). San Francisco: Jossey-Bass Publishers. Miles, M. B., & Hueberman, A. M. (1994). Qualitative data analysis: An expanded sourcebook (2nd ed.). Thousand Oaks, CA: Sage. Morse, J. M., & Richards, L. (2002). Readme first. Thousand Oaks, CA: Sage.
11 11 Patton, M. Q. (2002). Qualitative research & evaluation methods (3rd ed.). Thousand Oaks, CA: Sage. Plass, C., & Schetsche, M. (2000, December). The analysis and archiving of heterogeneous text documents: Using support of the computer program NUD*IST 4. Forum: Qualitative Social Research, 1(3). Retrieved from plassschetsche-e.htm Richards, L. (2002, February). Introducing NVivo: A workshop handbook (Rev. ed.). Doncaster, Victoria, Australia: Qualitative Solutions and Research. Richards, L. (1999). Using NVivo in qualitative research. Bundoora, Victoria, Australia: Qualitative Solutions and Research. Silverman, D. (2000). Doing qualitative research: a practical handbook. Thousand Oaks, CA: Sage. Silverman, D. (2001). Interpreting qualitative data: Methods for analyzing talk, text and interaction (2nd ed.). Thousand Oaks, CA: Sage. Taylor, S. J., & Bogdan, R. (1998). Introduction to qualitative research methods (3rd ed.). New York: John Wiley & Sons. Tesch, R. (1990). Qualitative research: Analysis types and software tools. New York: The Falmer Press.