CHAPTER 3: METHODOLOGY
3.7 Research design
3.7.5 Data analysis
This section describes the process of data analysis, including transcribing the interviews, managing the data, coding and the development of themes.
Qualitative analysis generally begins with a description of the data, the
generation of categories usually referred to as codes, and the identification of connections between codes to make interpretations (Leedy 1997; Grbich 2000; Padgett 1998). Conceptually, an interpretative approach ‘holds that people’s individual and collective thinking and action has a meaning which can be made intelligible’ (Minichiello et al. 1995: 248). It is described as a creative process requiring careful judgement about what is significant and meaningful in the data (Patton 1990). The aim of qualitative data analysis is:
…to find meaning in the information collected…systematically arranging and presenting information in order to search for ideas (Minichiello et al. 1995: 247).
In addition, Barritt suggests that qualitative data analysis is about trying to:
…go to the heart of the matter by looking for themes that lie concealed in the unexamined events of everyday life…to find meaningful, shared themes in different people’s descriptions of common experiences (Barritt 1986, cited in Leedy 1997: 162).
Data analytic techniques in qualitative research are often inductive, and include ‘juxtaposing emic accounts with codes, analysis and interpretation’ by the researcher (Padgett 1998: 110–111). Induction involves the development of generalisations from specific observations (Ruben & Babbie 1993: 54); an emic style of reporting includes first-person narratives that value the
respondents’ world view (Padgett 1998: 110–112). The approach taken in this study is primarily inductive and emic and, like Gregg (1994), applies a feminist method by giving credence to women’s voices and injecting into scholarly and activist discussions the voices, views, insights and experiences of women (Gregg, in Riessman 1994: 50).
Transcribing and managing the data
The audiotapes for each woman were transcribed verbatim onto a computer and averaged 30 pages in length. Each transcript consisted of thick
description and was analysed to identify the emerging central themes
(Riessman 1994; Minichiello et al. 1995; Padgett 1998). Each transcript was given a number and a pseudonym, with the latter attached to each unit of coded data. It was initially thought a qualitative data management computer programme, Ethnograph, would be a useful tool to both manage and assist with the process of analysing the large amount of data (approximately 1 600 pages). With only minimal training in its use, however, the technical task of entering the data and assigning codes became time-consuming and I began to perceive it as an onerous task. Months were spent entering and coding the transcripts using the Ethnograph programme.
However, working on the computer to analyse the data did not enable me to achieve the feeling of ‘being immersed in the data’ nor gain an overall, visual, ‘holistic feel’ for the data. In order to manage the data more effectively and intuitively, hard copies of each transcript were made and summary sheets devised for the cover of each transcript. From this developed a feeling of ‘being immersed in the data’ and a sense of creativity emerged (Patton 1990). Padgett (1998) suggests: ‘The goal of qualitative data management is to organise and store data for maximal efficiency in retrieval and analysis’
(Padgett 1998: 75). For this study, manually handling the approximately 1 600 pages of data became the most effective means of retrieval and analysis.
Coding and the development of themes
Based on the assumption within a feminist framework that women's perceptions of their experience of leaving and remaining out of a violent relationship with a male partner are meaningful and valid, codes and themes were identified (Riessman 1994; Minichiello et al. 2000; Padgett 1998). Content analysis involves a search for repeated key words, phrases,
sentences and themes. Some researchers count these elements in terms of frequency (Berg 1989), whereas Minichiello et al. suggest ‘these elements not only be counted…but…be examined for meaning’ (Minichiello et al. 2000: 252). As other researchers have found, discovering recurring themes takes the researcher months of reading and re-reading the transcripts and coding the data (Minichiello et al. 2000: 248).
Developing coding categories was initially achieved by highlighting, on hard copies of the text, the key words, phrases or concepts used by the
participants during the interview. This initial coding was based on categories matching the key topic areas in the semi-structured interview schedule that had been addressed by each participant (see Appendix 4).
The next phase of the analysis involved coding sub categories, the
development of additional categories and identifying similarities, contrasts and potential connections between them. This resulted in the generation of the
major themes. Using A5 sheets of paper, grids were initially used to visually present the codes which assisted with the creative process of developing new codes. The codes and themes that had emerged when coding the data in the Ethnograph programme were used as headings on the grids, resulting initially in 25 headings (new code and theme headings were added as they emerged from each re-reading of the transcripts). Each heading was then transferred to foolscap sized sheets of paper. Under each heading, the numbers one to 53 that were allocated to identify the participants’ transcripts in Ethnograph, together with their pseudonym, were written down on the left side of the page. Other information was included, for example age, region and number of children for easier cross-referencing throughout the analysis when identifying similarities, contrasts and potential connections. Brief summaries of the units of data in each transcript, related to the codes and themes, were then
recorded against the pseudonym, along with related short quotes and a page reference. Also included were page references to longer quotes that most effectively illustrated the codes and emerging themes, to be used for reporting of the results. When writing up the results this enabled ready retrieval of the data from the transcripts on computer.
Compiling these summaries under each new heading (code/theme) meant going through each of the 53 transcripts at least 15 to 25 times depending on the relevance of each heading to each participant. An example of a code or theme being relevant to every participant was ‘experiences of abuse and violence’. An example of when it may not have been relevant was codes related to having children where the participant did not have children. Having the pseudonyms attached to each unit of data throughout the entire analysis facilitated familiarity with the data and the development of linkages within and across the data. It also provided for a more holistic approach to the data, locating each woman within the emerging themes. It also helped me maintain respect for the uniqueness of each woman’s story whilst her experience was being generalised in the analysis of the data.
Once every participant’s contribution to each code had been summarised, an overall summary was compiled, enabling further development of emerging
themes and basic service usage data. This process resulted in 350
handwritten foolscap pages of data analysis. The combination of handwriting the coding, manually handling and keeping all the handwritten coding together facilitated a high level of immersion in the data and easier cross-referencing. This resulted in a high level of familiarity with the rich data. The visual
capacity to dip in and out of the data, and to spread it out in front of one-self, facilitated this process.
The use of pseudonyms, whilst time consuming, was an effective strategy aimed at maintaining a focus on the context of the data during the technical process of data analysis. It was also aimed at what I have described as reducing the ‘disembodiment’ of women from their quotes in the data
presentation. This strategy resulted in what I have termed ‘living’ data, that is, every code, category, theme, and quote evoked the women and the research context. Despite the relatively large sample for qualitative research (53), this facilitated the making of connections between and across data. However, as issues of confidentiality can arise when presenting data on recognisable groups (such as Aboriginal women or women from a culturally and
linguistically diverse background), quotes from women in these groups were not always accompanied by descriptions.
Throughout all phases of the analysis, the procedures of clarification,
confrontation and confirmation occurred. Previous feminist researchers have recorded the difficulties in relation to false consciousness versus researcher imposition (Acker, Barry & Esseveld 1983). They ask, how do we:
…produce an analysis which goes beyond the experience of the
researched while still granting them full subjectivity. How do we explain the lives of others without violating their reality? (Acker, Barry &
Esseveld 1983: 429).
However, Power refers to Hyde (1994) when she suggests that ‘limiting the research agenda to only giving voice would be an abdication of responsibility to link data from informants with broader theoretical frameworks to advance social change’ (Power 1998: 57). As part of addressing this issue, the
participants’ own words were a central focus of the findings, including: quotes from every woman; recording the specific information that each woman
wanted to share with other women (see Appendix 1); and providing a summary of each woman’s recommendations for improving formal and informal responses (see Appendix 2).
Avoiding imposition also requires the researcher’s reflexivity and self-critique (Riessman 1994; Padgett 1998; Darlington & Scott 2002). The process was assisted by the challenging data provided by several participants who reflected very different views to the majority of the participants and the researcher. The need to ensure that their views were incorporated into the analysis was a constant prompt for reflexivity as the data was analysed and emerging themes were identified. At the same time, connections between the women’s perceptions and relevant theoretical frameworks were made both in the analysis and the findings.
Methodological issues
Two methodological issues to emerge from the research were the inclusion of three participants who had left their partners earlier than the specified
timeframe. As each of these women belonged to a significant interest group (an Aboriginal woman, a woman from a CALD background and a woman who identified as having a disability), it was decided that their experiences could contribute to understanding the needs of each of these groups, albeit not specifically informing the research regarding more recent innovations and developments in practice and support in the domestic violence area.
The potential for sampling bias is acknowledged insofar as participants may not broadly represent women who have left a male partner who assaulted them and that the conclusions drawn can only be applicable to the
participants at the time of the study. Criticism of previous research in this area has been sample bias towards lower class groups who were more likely to come to the attention of helping agencies and police (Coorey 1988: 39). However, Coorey's research sample of 15 women who had experienced
domestic violence, obtained from police and court records, did result in a sample from different classes and backgrounds. McKinlay’s research sample of 15 women in Tasmania who had experienced domestic violence, obtained through networking and media publicity, also resulted in a sample from different socio-economic backgrounds (McKinlay [Patton] 1991). This research sample of 53 women resulted in a sample with a wide range of socio-economic backgrounds. It is not possible to know how representative the women are of the population in the general community.
As in the recent national study Against the Odds, this research included women who: had never told anyone about the violence; had only sought help from family and friends; used formal services but not domestic violence
services or police; used police and/or domestic violence services; and women who gained support from both formal and informal systems (Keys Young 1998: 7). The sample covers a range of women’s experiences to contribute to future policy and service development.
Validity and reliability
Many of the strategies for rigour in quantitative research, such as random sampling, generalisability and reliable and valid measurement, do not readily apply to qualitative research (Padgett 1998: 91). Rather, the goal of qualitative research is to evaluate trustworthiness and credibility using techniques like bracketing, searching for disconfirming evidence, triangulation, member checking, outlier analysis, chain of evidence, coding checks and the use of ‘thick’ description (Patton 1990; Leedy 1997). Other qualitative practices for validity include peer debriefing (Lincoln & Guba 1985), catalytic validity (Lather 1991); and construct validity where observations are consistent with current theory (Rubin & Babbie 1993). A number of these strategies were utilised to enhance trustworthiness and credibility.
The trustworthiness of the interview schedule was enhanced by both the piloting and consultation process, which was undertaken with social workers familiar with the topic, and academic researchers. To avoid interviewer
influence, consistency in tone of voice and phrasing of questions was aimed for in the application of each interview. Internal logic was aimed for by having one interviewer analyse all the interviews. Triangulation includes using
theories to check the validity of the findings; and if similar themes are noted in the data credibility is enhanced. The themes found in this study reflect
relevant theories identified in the literature. Peer debriefing was used both to critique the initial analysis of the data as the transcripts were coded and themes emerged, as well as throughout the study (Lincoln & Guba 1985). Member checking occurred when a number of participants were provided with the initial findings of the study through papers presented at numerous forums. Domestic violence service providers and government policy makers reviewed drafts of the research. The subsequent government report was published and provided to participants who also provided feedback.
Altheide and Johnson (1994) believe the usefulness, contextual
completeness, research positioning and reporting style of a study (referred to as interpretative validity) can also be used to judge the validity of qualitative research (Altheide & Johnson 1994, cited in Leedy 1997: 168). Leedy states:
Usefulness refers to whether the research report enlightens those who read it or moves those who were studied to action. Contextual
completeness refers to the extent to which a comprehensive review of the situation is provided (Leedy 1997: 168).
Leedy suggests completeness can be obtained by providing information about both the history of the topic studied and participants’ perceptions and
meanings (Leedy 1997: 168). Research positioning requires the researcher to be both aware of their own influences and to make them explicit ‘so that readers can determine for themselves the credibility of the findings’ (Leedy 1997: 168). This is enhanced by the researcher’s reflexivity: ‘the continued consideration of the ways in which the researcher’s own social identity and values affect the data gathered and the social world produced’ (Reay 1996: 60). In relation to a researcher’s reporting style, ‘the reconstruction of participants’ perceptions must be perceived to be authentic’ (Leedy 1997: 168). According to each of these ‘measures’, the feedback on the government
report, provided by participants, suggests the study has interpretative validity.
There is also an assumption inherent in the subjective paradigm of qualitative research that participants will be able to reflect on their experiences and accurately express their perceptions of those experiences. A possible extraneous factor may be that a participant’s recall can be a limitation, affected by the time that has elapsed and recent history. To minimise the elapse of time, the sample was taken from women who had left their partners as recently as possible without impinging on situations where interventions were current. The minimum time that had elapsed was usually two years; and in the majority of cases, the maximum time was five years. Participants who had indicated at the time of interview their desire to receive copies of their transcripts were sent copies and invited to correct any transcribing errors, amend their recorded responses or make further comment.
3.8
Summary
This chapterhas discussed the methodology of the study and the method utilised, describing a qualitative (exploratory and descriptive) study conducted from a feminist standpoint (Hartsock, in Harding 1987; Bell & Klein 1992; Grbich 1999). The research aims and questions were outlined and a definition of key terms provided. The researcher’s conceptual framework was
discussed, identifying the influences of feminist and structural social work theory on the methodology (Reinharz 1992; Kelly et al. 1994; Mullaly 2002). The choice to utilise qualitative research methods was justified, supporting an analysis of social structures and policies and suitable for the sensitive nature of the topic and the exploratory nature of the research (Oakley 1988;
Riessman 1994; Padgett 1998; Grbich 2000). The research design was described, including ethical considerations, the interview schedule, how the sample was obtained, methodological issues arising, the interview process, the data analysis and the measures used to enhance trustworthiness and credibility. The following chapters report the findings of this study.