OVERVIEW OF CHAPTER TWO: METHOD
2.4 DATA ANALYSIS
This section describes how quantitative and qualitative data were analysed, with particular emphasis on describing the grounded theory approach to data analysis.
Analysis of quantitative data
Quantitative questionnaire responses were analysed using the ‘SPSS for Windows’ statistical package.
Chapter Two: M ethod
Analysis of qualitative data
Qualitative data were obtained from the responses to the open-ended questions in the questionnaire, and from the focus group discussions.
Data were collected from questions 42, 56-57, 68-72 in the questionnaire. These questions asked about the ways in which the group had helped (Q.42), what had been most helpful (Q.56) and unhelpful (Q.57) in group meetings, what difference being a member had made (Q.68), whether respondents had changed since joining the group, and if so how (Q.69), how things would be different without the group (Q.70), how respondents would recommend the group to another (Q.71), and an opportunity for people to say anything else about their experiences in the group (Q.72).
Obviously these questions were asking about different aspects of people’s
experiences, and asked about both processes and outcomes. Nevertheless, replies to all of these questions were analysed together to draw out the key features of people’s experiences. Previous researchers have also recommended that the distinction
traditionally made in treatment evaluation between process and outcome is not appropriate for mutual support group research (e.g. Humphreys & Rappaport, 1994; Levy, 1979; Lieberman & Borman, 1976).
Each focus group discussion was tape recorded, and field notes of each discussion were taken as soon afterwards as possible, as recommended in the literature (e.g. Morgan, 1988). Each discussion was transcribed by hand.
Data from the questionnaires and focus group discussions were each analysed separately but in the same way, using the grounded theory approach to qualitative data analysis and the constant comparison method (Glaser & Strauss, 1967; Henwood & Pidgeon, 1992; Pidgeon, 1996; Pidgeon & Henwood, 1996; Strauss & Corbin,
1990).
Grounded theory approach and method
The key analytical method in the grounded theory approach is constant comparative analysis. Here the researcher is urged to be constantly alert to the similarities and differences which exist between instances, cases and concepts, to ensure that the full diversity and complexity of the data is explored (Henwood & Pidgeon, 1992; Pidgeon & Henwood, 1996; Strauss & Corbin, 1990). The process of open coding (Strauss & Corbin, 1990) is one in which, as categories emerge, units of analysis are compared to each category. Each unit is assigned to as many categories as possible. If no
categories fit a given unit, a new category is developed to represent it. After the categories have been saturated (i.e. no new categories emerge from further data), the analytic focus shifts to the relationships among the categories. The process of axial coding (Strauss & Corbin, 1990) compares open categories and it becomes apparent that the network of linked categories forms a hierarchical structure in which central categories subsume lower-order categories. This process may continue until a single, core category emerges, which is most densely related to the other categories. It is
Chapter Two: Method
typically an abstract category but it is not vague (Pidgeon, 1996; Rennie, Phillips & Quartaro, 1988; Strauss & Corbin, 1990).
In this study both sets of qualitative data were inspected and broken down into meaning units. These units were then assigned to categories according to the process of open coding (Strauss & Corbin, 1990). Categories were created to fit the
emerging data until saturation point was reached, namely the point when the data were assigned to existing categories without the need for further categories to be created. In each case this stage produced a large number (running into hundreds) of first order categories which closely fitted the data. These were then clustered into a number of second-order categories, often labelled with participants’ own words. The frequencies with which each of these second-order categories were mentioned in the questionnaire responses were recorded and are reported in Chapter Three. This adds valuable detail to the analysis by giving an indication of the relative importance
attached to each category by the participants. Silverman advocates an approach which includes basic quantitative analysis with qualitative studies by arguing that ‘simple counting techniques can offer a means to survey the whole corpus of data ordinarily lost in intensive qualitative research’ (1985, p. 140).
The relationships between these second-order categories were then examined according to the process of axial coding (Strauss & Corbin, 1990), to create a hierarchy of categories. In each case it was possible for a single core category to emerge, which was densely related to the lower-order categories.
The focus group data were analysed after the qualitative questionnaire data. It was appropriate to analyse these data from a different source separately, because they were spoken rather than written, because people were responding to different questions, and because they emerged from group discussion rather than purely individual responses. This analysis was undertaken after the analysis of the
questionnaire responses, and so is open to the possible criticism that it was influenced by the previous analysis. The lower-order categories which emerged in the analysis of the focus group data were different to those in the analysis of the questionnaire data. However, in each case the same three higher-order categories and single core
category emerged consistently.
Another way of addressing the issue of reliability is replication. In this study, the axial coding process (Strauss & Corbin, 1990), in which the first-order categories were grouped into higher-order categories, was replicated by another researcher
experienced in the grounded theory approach and method. Labels and definitions of the first-order categories were provided to this second researcher, who then
undertook the axial coding process. There were some differences in the assignment of a small number of individual first-order categories. However, the same three higher- order categories emerged from this replication. This supported the most central findings of the analysis in terms of the higher-order categories, while providing a reminder that in some cases the boundaries between those categories were slightly blurred or overlapping.
Chapter Two: M ethod
After the data were analysed, a summary report was sent to Contact a Family and to all participating groups. Discussions with Contact a Family constituted a form of respondent validation, in which the research findings were recognised by staff and advisors in the mutual support organisation as corresponding broadly with their experience.