Research Issues and Methods: An Introduction for Educational Technologists
40.3 ANALYZING QUALITATIVE DATA
Qualitative data are considered to be the "rough materials researchers collect from the world they are studying; they are the particulars that form the basis of analysis" (Bogdan & Biklen, 1992, p. 106). As described earlier, qualitative data can take many forms, such as photos, objects, patterns of choices in computer materials, videotapes of behaviors, etc. However, words often are the raw materials that qualitative researchers analyze, and much advice from researchers discusses analyzing these words.
The need for brevity in this chapter precludes an extensive discussion of analyzing qualitative data. However, we will introduce the researcher to the issues underlying decisions to be made and provide several views of how to analyze data. As noted by Miles and Huberman (1994) in their in-depth sourcebook, beginning researchers may quake in the face of the "deep, dark question" regarding how to have confidence that their approach to analysis is the right one (p. 2). Yet we concur with the thoughtful and yet practical approach of these authors, that one must just begin and that more energy is often spent discussing analysis, and research for that matter, than "doing it." Miles and Huberman note, in a decidedly unnaive approach, that "...any method that works, that will produce clear, verifiable, credible meanings from a set of qualitative data," is "grist for their mill. They add, "...the creation, testing, and revision of simple, practical, and effective analysis methods remain the highest priority of qualitative researchers," adding that, "We remain convinced that concrete, shareable methods do indeed belong to 'all of us"' (p. 3). It is in this spirit that we present approaches to analyzing qualitative data.
One of the major hallmarks of conducting qualitative research is that data are analyzed continually, throughout the study, from conceptualization through the entire data collection phase, into the interpretation and writing phases. In fact, Goetz and LeCompte (1984) describe the processes of analyzing and writing together in what they call analysis and interpretation. How these activities may be done will be explored below.
40.3.1 Overall Approaches to Analyzing Qualitative Data
Qualitative researchers choose their analysis methods not only by the research questions and types of data collected but also based on the philosophical approach underlying the study. For example, Miles and Huberman (1994) outline three overall approaches to analyzing qualitative data. An "interpretive" approach would be phenoniological in nature or based on social interactionism. Researchers using this
approach would seek to present a holistic view of data rather than a condensed view. They might seek to describe a picture of "what is." They would generally not choose to categorize data to reduce it. Miles and Huberman note that the interpretive approach might be used by qualitative researchers in semiotics, deconstructivism, aesthetic criticism, ethnomethodology, and hermeneutics.
A second approach described by these researchers is "collaborative social research," often used by action researchers in partnerships composed of members of many, and sometimes opposing, organizations.
The final approach to analyzing data described by Miles and Huberman is that of "social anthropology," which relies primarily on ethnography. Researchers using this approach seek to provide detailed, or rich, descriptions across multiple data sources. They seek regular patterns of human behavior in data, usually sifting, coding, and sorting data as they are collected, and following up analyses with ongoing observations and interviews to explore and refine these patterns, in what Goetz and LeCompte call a recursive approach (1994). Researchers using a social anthropology approach also tend to be concerned with developing and testing theory. Researchers who develop life histories, work in grounded theory and ecological psychology, and develop narrative studies, applied studies, and case studies often base their analyses on this social anthropology approach. Many of the methods for, and views about, analyzing qualitative data can be seen to be based on this social anthropology approach.
40.3.2 Methods for Analyzing Qualitative Data
Depending on the basic philosophical approach of the qualitative researcher, many methods exist for analyzing data. Miles and Huberman state that qualitative data analysis consists of "three concurrent flows of activity: data reduction, data display, and conclusion drawing/verification" (1994, p. 10). Most researchers advocate that reducing and condensing data, and thereby beginning to seek meaning, should begin as the study begins and continue throughout data collection.
22.214.171.124. Data Reduction. Goetz and LeCompte (1994) describe the conceptual basis for reducing and condensing data in this ongoing style as the study progresses. The researcher theorizes as the study begins and builds and tests theories based on observed patterns in data continually. Researchers compare, aggregate, contrast, sort, and order data. These authors note that while large amounts of raw data are collected, the researcher may examine in detail selected cases or negative cases to test theory. They describe analytic procedures researchers use to determine what the data mean. These procedures involve looking for patterns, links, and relationships. In contrast to experimental research, the qualitative researcher engages in speculation while looking for meaning in data; this speculation will lead the researcher to make new observations, conduct new interviews, and look more deeply for new patterns in this "recursive" process.
Researchers may derive patterns in many ways. They may, for example, engage in what Goetz and LeCompte call "analytic induction" (p. 179), reviewing data for categories of phenomena, defining sets of relationships, developing hypotheses, collecting more data, and refining hypotheses accordingly. As noted earlier, interpretivists would be unlikely to use this method. They would not tend to categorize, but would scan for patterns in order to build a picture or tell a story to describe what is occurring.
Another method, constant comparison, would be relied on by those using a grounded-theory approach. This method involves categorizing, or coding, data as they are collected, and continually examining data for examples of similar cases and patterns. Data collection can cease when few or no new categories of data are being encountered. Goetz and LeCompte contend that researchers using constant-comparison code data look for patterns as do those using analytic induction, but the categories are thus processed differently.
Bogdan and Biklen (1992) describe in detail practical approaches to writing up field notes, one of the main forms the "words" that make up qualitative data take. They recommend writing field notes with large margins in which to write later notes as data are later analyzed, as well as in which to write codes for these data. They also advise that text be written in blocks with room left for headings, notes, and codes.
It should be noted that virtually all researchers who use an ethnographic approach advocate writing up field notes immediately after leaving the research site each day. Observations not recorded will quickly be forgotten. Researchers may not realize the importance of some small phenomenon early on, so these details should be recorded each day. Most authors further recommend that researchers scan these data daily, analyzing thoughtfully for patterns and relationships, and perhaps adding to or modifying data collection procedures accordingly.
Field notes consist of observations and the researcher's interpretations. Bogdan and Biklen (1984) call these two types of field notes contents the descriptive part (p. 108) and the reflective part (p. 12 1). They state that the descriptive part consists of detailed descriptions of the subjects and settings, the actual dialogue of participants, descriptions of events and activities, as well as descriptions of the observer's behavior, to enable determining how this may have influenced participants' behaviors. The reflective part of field notes, they add, consists of the observer/researcher Is analysis. The researcher records speculations about patterns and how data can be analyzed, thoughts about methods and ethical concerns, and even ideas about his or her own state of mind at the time. These ' authors provide many pages of actual field notes from studies done in elementary and secondary education classrooms, which the beginning researcher will find helpful.
If researchers collect data using audiotape or videotape, written transcripts of language recorded are often prepared. Later analysis can be done, but notes should still be recorded immediately after being in the field. Such notes, for instance, will include observations about participants' nonverbal behaviors, what was occurring in the immediate surroundings, or what activities participants were engaging in. Even in the case of interviews, notes might include these descriptions, as well as what participants were doing just prior to interviews. As noted in the discussion of data collection methods, audiotapes and videotapes may be subjected to detailed microanalysis. Usually data are coded and counted, but, due to the labor-intensive nature of this type of analysis, segments of these "streams of behavior" are often systematically selected for analysis.
It is advisable to collect data in its raw, detailed form and then record patterns. This enables the researcher later to analyze the original data in different ways, perhaps to answer deeper questions than originally conceived. The researcher many weeks into data collection may realize, for example, that some phenomena previously considered unimportant hold the keys to explaining participants' views and actions. In addition, preserving the raw data allows other researchers to explore and verify the data and the interpretations.
If researchers have collected documents from subjects, such as logs, journals, diaries, memos, and letters, these can also be analyzed as raw data. Similarly, official documents of an organization can be subjected to analysis.
Collecting data in the form of photographs, films, and videotapes, either those produced by participants or by the researcher, has a long tradition in anthropology and education. These data, too, can be analyzed for meaning. (See, for instance, Bellman & Jules-Rosette, 1977, A Paradigm for Looking; Bogdan & Biklen, 1992; Bogaart & Ketelaar, 1983, Methodology in Anthropological Filmmaking; Col-lier, 1967, and Collier & Collier, 1986, Visual Anthropology as a Research Method; Heider, 1976, Ethnographic Film; and Hockings, 1975, Principles of Visual Anthropology.)
40-3.2.2. Coding Data. Early in the study, the researcher will begin to scan recorded data and to develop categories of phenomena. These categories are usually called codes. They enable the researcher to manage data by labeling, storing, and retrieving it according to the codes. Of course, the codes created depend on the study, setting, participants, and research questions, because the codes are the researchers' way of beginning to get at the meaning of the data. There are therefore as many coding schemes as researchers. Still, examples of coding schemes will here be provided in an attempt to guide the reader.
Miles and Huberman (1994) suggest that data can be coded descriptively or interpretively. Unlike some authors, they suggest creating an initial "start list" (p. 58) of codes and refining these in the field. Researchers using a strictly inductive approach might choose not to create any codes until some observations and informal interviews were conducted from which codes could be induced.
Bogdan and Biklen (1992) recommend reading data over at least several times in order to begin to develop a coding scheme. They describe coding data according to categories and details of settings; types of situation observed; perspectives and views of subjects of all manner of phenomena and objects; processes, activities, events, strategies, and methods observed; and social relationships. Goetz and LeCompte (1984) describe coding to form a taxonomic analysis, a sort of outline of what is related to what, and in what ways.
In one of many examples he provides, Spradley (1979) describes in extensive detail how to code and analyze interview data, which are semantic data as are most qualitative data. He describes how to construct domain, structural, taxonomic, and componential analyses. We will discuss, as one example, domain analysis. Domains are names of things. Spradley proposes "universal semantic relationships" that include such categories as "strict inclusion," that is, "X is a kind of Y'; "spatial," "X is a place in Y, X is a part of Y'; and "cause-effect," "rationale," "location of action," "function," "means-end," "sequence," and "attribution" (P. I 11). Spradley provides an example from his own research. In a study on tramps, he found from interviews that the cover term flop, as a place to sleep, included such things as box cars, laundromats, hotel lobbies, and alleys.
An example of the types of codes that might be developed to investigate patterns of teacher use of an educational technology innovation is presented in the Savenye and Strand observational study described earlier (1989). The researchers videotaped teachers and students using the multimedia science course in 13 physical science classrooms in four states. Samples of videotapes from three teachers were selected for approximate equivalence; in the samples, the teachers were teaching approximately the same content using the same types of lesson components. The researchers were interested not in all the behaviors occurring in the classrooms but in the types of language expressed as teachers taught the lessons.
After reviewing the videotaped data several times, the researchers developed codes for categorizing teacher language. Most of these codes were created specifically for this study. For example, the most frequent types of teacher language observed were instances of "teacher statements," which included data coded as "increasing clarity or coherence of information presented." Examples of codes in this category included PR, for providing preview or organizers of lessons; RP, reminding students to remember prior knowledge; EL, elaborating by providing new information about a scientific concept in the lesson; and R, providing a review of lesson content. Another example of a code created for teacher statements was REL, for instances of when a teacher relates content to student's own experience with everyday examples.
Savenye and Strand were also interested in the types of questions teachers added to the curriculum to encourage their students to participate actively during the whole-class presentations of content. Along with a few created codes, the researchers developed codes based on Bloom's taxonomy of cognitive objectives (1984). Such codes included REC, for questions that asked students to recall information just presented by the multimedia system; APP, for questions that required students to apply or extend lesson content to new content or situations; and ANAL/SYN, for questions that require a student to analyze a situation to come up with solutions or to synthesize a solution. In a result similar to those of many studies of teacher-questioning strategies, but which may disappoint multimedia developers, the majority of the teachers' questions simply asked students to recall information just presented, rather than to apply or analyze or synthesize knowledge learned.
In this study, as in most qualitative studies, coding schemes were continually added to, collapsed, and refined as the study progressed. However, in some studies, only preassigned codes are used to collect and/or analyze data. As in the use of Bloom's categories by Savenye and Strand (1989), usually these codes have been derived from studies and theories of other researchers, or from pilot studies conducted by the researchers themselves. These studies may use observational coding forms or protocols on which data are recorded in the coding categories.
Another example of using preassigned codes is a study conducted to investigate how visitors to a botanical garden use interactive signs (Savenye, Socolofsky, Greenhouse & Cutler, 1995). Among other types of data collected in this study, these researchers trained observers to record behaviors visitors engaged in while they used signs. Observers recorded whether visitors stopped to read a sign at all; if so, for how long, and the level of interactivity visitors exhibited. Based on the work of Bitgood (1990), interactivity was coded as stopping briefly and glancing only; obviously reading the sign and looking at the plant exhibit near it; and, finally, engaging in highly active behaviors, such as reading the sign aloud, pointing to the plants displayed, discussing information being learned, and pulling friends and family over to the sign to read it. In a blend of coding methods typical in many studies, observers also wrote ethnographic-style notes to-describe what if any content-on the signs was being discussed, what misconceptions appeared, what excited visitors most, etc. In this study, visitor surveys and interviews were also used.
In any qualitative study, codes can be used to count frequencies or, as Goetz and LeCompte call it, conduct enumeration (1984) to develop quantitative data, as was done in the studies described above. Similarly, quantitative data, such as attendance or production figures, from other sources, may be analyzed. Most researchers suggest caution when counting that the "big picture" is not lost, and also note that quantitative data from other sources can also be biased. Even what is collected in a school district, for instance, may be determined by financial, administrative, and political concerns.
For more examples of coding schemes and strategies, see Strauss (1987). (See also 41.2.4 for some discussion of coding observational data.)
126.96.36.199. Data Management
188.8.131.52. 1. Physically Organizing Data. Analysis of data requires examining, sorting, and reexamining data continually. Qualitative researchers use many means to organize, retrieve, and analyze their data. Many researchers simply use notebooks and boxes of paper. Bogdan and Biklen (1992) describe what they call two mechanical means to organize and begin to review data. One way they describe is to write initial codes in margins of field notes, photocopy the notes and store the originals, then cut up and sort the text segments into piles according to codes. These coded data can be stored in boxes and resorted and analyzed on an ongoing basis. A second method they describe is to record field notes on pages on which each line is numbered, code the field notes, and then write the page number, line numbers, and a brief description of each piece of data on a small index card. These cards can then be sorted and analyzed. The authors note that this second method is better suited for small sets of data, as it often requires returning to the original field notes to analyze the actual data.
184.108.40.206. Organizing Data Using Computers. Computers are increasingly the tool of choice for managing and analyzing qualitative data. It is interesting to note that computers have long been used in anthropological analysis. (See, for instance, Hymes, 1965, The Use of Computers in Anthropology.) Computers may be used simply for word processing in developing field notes. However, there is now considerable software specifically developed for qualitative research, and it can be expected that many new programs will be developed in the upcoming decade. Some software uses text entered with a word processor to retrieve words and phrases or to manage text in databases. Software is also available to code and retrieve data, and some programs also allow for building theories and conceptual networks. Programs are available for IBM (e.g., QUALPRO, The Ethnograph) or for Macintosh microcomputers (e.g., HyperQual, SemNet) or multiple systems (QSR NUD-IST) (Miles & Weitzman, 1994). For much more on using computers for analysis, the reader may refer to the following books: Tesch (1990), Qualitative Research: Analysis Types and Software Tools, or Wietzman and Miles (1995), A Software Sourcebook: Computer Programs for Qualitative Data Analysis.
220.127.116.11.3. Data Display. Seeking the meaning in data is made easier by displaying data visually. Research data are displayed using charts, graphs, diagrams, tables, matrices, and any other devices, such as drawings, that researchers devise. Frequency tables are typically developed for categories of coded behaviors. In the Reiser and Mory (1991) study, for example, teachers' planning behaviors were coded and tables of behaviors presented.
Miles and Huberman (1994) hold that data display is a critical and often underutilized means of analysis. They describe many forms of data display, illustrated with examples of actual data. They recommend that researchers initially create categories of data, code data, and revise codes, as do other authors. They note that increasingly qualitative research involves analyzing what they call within-case data, for instance, from one classroom or one school, as well as "cross-ease" data, from many participants and many sites. Whereas in one case study, it may not be necessary to present visual displays-narrative descriptions might suffice - studies involving data from many cases can greatly benefit from visual displays. Miles and Huberman present many options. For example, for within-case data they show context charts or checklist matrices, but they also discuss using a transcript as a poem. They also illustrate time-ordered displays, role-ordered displays, and conceptually ordered displays. For cross-case studies, these researchers mention some of the earlier displays for reviewing and presenting data, along with case-ordered displays. They illustrate other displays for examining cross-case data and provide extensive advice for creating matrix displays.
An example of the use of matrix displays is the Higgins and Rice participant observation study described earlier (1991). The researchers analyzed teachers' conceptions of all the activities that represent "assessment." These data were derived from a series of structured interviews with the teachers, conducted in conjunction with observations of the teachers and their students. The researchers analyzed these data using multidimensional scaling and displayed the data using a matrix to show the relationships among types of assessments teachers used and how different teachers conceived of them differently.
That data analysis is woven into interpreting results and writing up the study is indicated by the fact that Miles and Hubetinan describe the third type of data analysis activity as drawing and verifying conclusions. Similarly, Goetz and LeCompte (1984) include writing up the study in their chapter on analysis and interpretation of data, describing the writing phase as developing an ethnographic analysis, and integrating and interpreting the study. While recognizing that analysis continues as the research report is written, and that writing should begin during analysis, in this chapter, we will present ideas and issues for writing up a study.