Missing data: finding 'central' themes in qualitative research

Qualitative sociology relies heavily on the "emergence" of themes, construction of categories "out of" data, and linking those categories to form theories. But the processes by which the researcher constructs theory are rarely reported in detail. The literature is clear on the goals of theme emergence and theory construction, but not nearly so clear on the ways their relationships to data are established, or the processes by which categories can be linked to and tested through, let alone "grounded" in, data. Researchers can remain untroubled by these issues, and reports often do not mention doubts or dilemmas attending the process of theory construction. This paper offers a contribution to the practical consideration of theory construction processes, and particularly to the practice of exploring the complex and uneven data records that usually confront the researcher. It argues that "core" categories are quite normally not evident in data, and that the gaps in data may be more significant than areas that are well documented. We report two experiences in unrelated projects of two kinds of “missing data”. The first relates to the emergence of grounded theory, in which the categories that proved central to the analysis were hardly represented in the "raw" data. We trail the processes by which the researchers became increasingly convinced of the centrality of these "missing" themes, and set about establishing an explanation for their failure to appear in the open-ended accounts that formed the data. We then raise a question not often asked: how to establish the validity of the explanation for which there is at first sight no evidence? In the second type of missing data, the data are missing, because the project raised questions at a late stage which the data did not directly address. Changes in the research question are common. However when the question emerges towards the end, especially of a funded project, then it becomes the starting point of a new project. Introduction The literature on qualitative research has focused on ways researchers make sense of data through coding, memoing and writing. The assumption is that the data are there. But the experience of theory construction is more often one of building jigsaws from pieces of data 2 and observation in studies where data take many forms and the researcheru0027s interpretation is itself data. In that setting, our experience is that quite normally the jigsaw will have missing pieces, whose "missingness" is the key to understanding. This paper resulted from discussions with novice and experienced researchers concerned at how qualitative research literature portrays the relationship between theory and data. It is rare to find research accounts that do not make the emergence of theory appear a smooth, even inevitable process. Our own experiences, and those of our students, have never fitted such smooth images, and in discussions we have often found that others are helped by our accounts of the puzzles and anxieties, and the hard detective work, which we have experienced during the analysis stage when a picture appeared to be emerging, but jigsaw pieces were evidently missing. Missing data are seldom reported in accounts of the research process. To discover that data do not adequately cover the subjects required is thus often interpreted as a personal failure. The doubts are manifold, revolving around the appropriateness of the research question, the data collection methods, the analysis of data and ultimately the researcher’s ability to do research. In this paper we argue, on the contrary, that in qualitative research it is inevitable there will be missing data, for two different sets of reasons. Firstly, missing data is always an issue since the data collection processes are not designed in advance to answer a known question. In a grounded study where the aim is to discover theory from data (Glaser u0026 Strauss, 1967) (Strauss u0026 Corbin, 1990), the theory will always require more data, and the processes of "theoretical sampling" are always only partial. The data are missing even when researchers collect and analyze data concurrently so as to be able to adapt interviews to new questions as they emerge (Miles u0026 Huberman, 1984). A common experience in qualitative work is that the research question evolves as 3 understanding grows. Thus the question at the end of the study differs from the one that initiated the study. The gaps in the data thus emerge when the initial data are viewed through the lens of this subsequent question. This is not a matter for concern. Such a change in the research question is a normal aspect of research, indicating progress in understanding. Popper (1972) argues that the change in question is an essential characteristic of the growth of knowledge. All knowledge grows by an initial problem generating a tentative theory. Critical discussion leads to a process of error elimination which in turn generates the next problem. Secondly, since meanings and interpretations are the stuff of qualitative studies, data are always missing because there are some questions that respondents do not want to answer, others they circumvent, and most significantly, aspects of their experiences are so taken for granted that they are not reported. In such situations, the silence on these issues in itself becomes powerful data, and the aware researcher can focus on such silences as input to the processes by which they generate theory. It is the “missingness” of data which often provides the most significant theoretical challenges. The more massive the missingness of data, the more likely it is going to be a major theoretical indicator. Grounded theory has welltested ways to deal with a change of question. Researchers collect and analyse data at the same time, so as to keep redirecting the research to emerging themes. Theoretical sampling, where data are collected to test theory and the “validity of findings” (Finch u0026 Mason, 1990), is often useful. The u0027missingnessu0027 problem arises when new questions come up so late in the study that further collection or recoding of data is not possible. The fit of data and theory (Glaser, 1978), an essential part of grounded theory, is often found wanting during the final writing up stage of research. The scope for recoding or re-collecting data is particularly limited in funded studies where external guidelines have to