Data Analysis

Analytic decisions are generally made based on the research questions, the study model, and the types of data collected. Analysis of qualitative data progresses through classification of ideas, themes, topics, activities, types of people, and other categories relevant to the study. This process is referred to as coding. Coding involves the classification of elements in text data into categories that are related to the study topic and are useful in analysis. Corbin and Strauss (2008) discuss different types of coding categories, ranging from more concrete to more abstract and conceptual. These coding categories are created and refined as the researcher builds smaller units into larger domains. The researcher examines and describes variation within each code category, identifies links among code categories, tests these with further examples, and then explains and interprets them (see Chapter Three, Grounded Theory). As analytic codes emerge or initial codes are applied, a coding scheme is finalized that can be applied to the entire data set. All text-based observation and interview data can be coded, compared, and integrated into patterns by hand, or by using computer software, such as Atlas-ti or NVivo (QSR International,; Scientific Software, 2010). Furthermore, many of these programs allow for the incorporation of audiovisual, photographic, PDF, and JPG formats into files that can be coded.

Either with or without software, analysts make comments about interesting points or codes, and develop thematic, theoretical, methodological, or other types of memos than can be analyzed along with the data. Good qualitative researchers also write analytic summaries that provide the basis for overall project analysis and interpretation. An analytic summary is an interim write-up of the results of close examination of a specific coding category, for example, “partner relationship.” To write a summary, the researcher would extract everything that has been coded or classified as about partner relationship. This could include different types of partners and different dimensions of relationships. An analytic report on this component of a study would sort out the different types of relationships and consider under each type the different dimensions of the relationship. The write-up would then summarize characteristics of relationships for each partner type and synthesize similarities and differences in relationships across all partner types. Chapter Three includes additional information about the process of data analysis.

Conceptual mapping, network research, drawings and photographs, geographic mapping, and other advanced qualitative methods or tools each require a different approach to analysis (LeCompte & Schensul, 2010; Miles & Huberman, 1994) and different software. Conceptual mapping often makes use of Anthropac, a program written by a sociologist for social scientists to facilitate free listing and grouping of items in a single domain (Analytic Technologies, 2010). Network analysis can be conducted with SPSS (for personal networks), and such programs as UCINET (for macro-network analysis) and Pajek or Krackplot (for macro-network display) are available. Many anthropologists use Microsoft ACCESS and GIS mapping software to create community maps or personal geographic spaces, which can then be compared and contrasted for differences across individuals, neighborhoods, or communities and villages.

To interpret their data, qualitative researchers triangulate different types of data, comparing and contrasting results to find and explain commonalities and differences. Triangulation refers to an examination of how different sources of data on the same topic may complement each other to deepen understanding of a study topic. For example, community-level interviews on perceptions of alcohol use among men in Mumbai can be complemented by consensus analysis that reveals the way men and women classify reasons men drink and the activities in which men are engaged (Berg et al., 2010; Schensul et al., 2010). The community-level interviews with local experts have shown the widespread belief that many men drink and that drinking leads to risky sexual activity. Consensus analysis showed that both men and women classified reasons for drinking into primary categories related to work conditions, stresses and conflict at home, influence of alcohol-consuming peers, parties and social life, and risky activities, such as going to ladies' bars to find women and seeing female sex workers. Consensus analysis broadens and complements the perspective provided by a much smaller number of local respondents.