Qualitative data: coding and analysis
The following text provides information on the techniques most commonly used in analysing data collected by means of focus groups, interviews and ethnography (the techniques most likely to be used in qualitative studies of new food technology development). Emphasis is placed on coding and thematic analysis. Other techniques such as discourse analysis are not included. Qualitative data consists predominantly of text collected using methods such as interviews and focus groups (generally transcribed from an audio recording) and field notes from observational studies. Other data types exist e.g. video footage (which may lead to quantitative or qualitative analysis.)
Several methods exist for qualitative analysis and each has underlying procedures and assumptions. However, unlike quantitative analysis, there is not an established, unambiguous set of rules for the analysis of quantitative data (Bryman 2012: 566).
Analysis may be inductive or deductive. The inductive approach is used when there is little previous knowledge. The framework for analysis (concepts, categories) is drawn out of the text by the researcher. A particular inductive approach is Grounded Theory. It lays down strict procedures for the conduct of the analysis with the object of drawing out theory from data. Initially developed by Glaser and Strauss (1967) it has been hugely influential in qualitative research. Although researchers very frequently use important elements of the approach, it is much rarer to find ‘pure’ examples of its application. Another approach is the generic inductive qualitative model (Bryman, 2012, p 422). The deductive approach is used when the purpose of the study is to test existing theory, hypotheses, concepts or models, and the framework for analysis (categories etc) is derived a priori from existing literature.
Qualitative textual analysis requires repeated detailed reading of the text and then, following a systematic and transparent process, the reduction of the mass of data (possibly covering hundreds of pages) to a briefer format. This process begins with coding, by which small units in the text which are potential indicators of concepts are identified.
Main steps in analysing text
Coding Coding is the starting point for much qualitative analysis and is a process by which the very large amounts of data (text generated from interview, focus group, notes from field observations etc) can be reduced to manageable units for analysis. By following the process the analyst becomes immersed in the data.
Computer-assisted qualitative data analysis (CAQDA)
In many respects CAQDA packages (such as NVivo) perform the same processes that are done manually – code and retrieve. The researcher can code the text, create categories (nodes) and concepts while working at the computer. They can also generate memos relating to these analytical components. Clearly the interpretative skills of the researcher are still required. They can then very quickly retrieve text sequences relating to any code, category or concept. This can greatly reduce the time spent on data-handling during analysis. They also provide an easy means of visualising (representing) the relationship between codes, categories and concepts.
Thematic analysis This is a very commonly used method in qualitative data analysis and is applied to text generated from interviews, focus groups etc. It is a flexible, widely applied approach and does not have a single well-defined set of procedures. After repeated reading, the data are coded, and the analyst looks for categories or themes and subthemes which recur that are relevant to understanding the topic being researched. Themes may be inspired by theoretical ideas, or they can emerge from the data. A framework is created which displays the themes (in columns), the individual interviews (rows), and, in the cells, fragments of text showing the occurrences of the theme each interview. Thus comparisons and contrasts can be made between different respondents for each theme and also theme-by-theme comparisons can be made. Different frameworks can be developed for the same text. The whole set of interviews is analysed without further data collection.