CHAPTER FOUR: RESEARCH METHODOLOGY AND METHODS
4.11 Data Analysis process
4.11.7 Managing and Analysing Questionnaire Data
A number of different ‘schools of thought’ stress that there is no single kind of qualitative data analysis or application but a range of approaches and procedures related to the data to be analysed, different perspectives, purposes and preferences of the researchers (Dey, 2005, p2). Several analytic strategies can be used in qualitative data analysis, including grounded theory, narrative and discourse analysis. Researchers can also opt to analyse data the old fashioned-way by cutting and pasting pieces of paper or using computer-based analysis programs (Mertens, 1998). Some of the general features common to the analytical phase of qualitative research were followed and these include:
161
General features common to the analytical phase of qualitative research
Some form of review of all the information to gain an initial insight of the data.
The process of organising data into some manageable form which is often referred to as ‘reducing data’ and typically involves developing codes and/or categories.
Interpreting data.
Presenting it diagrammatically in different forms (Mertens, 2010; Dey, 2005; Ritchie and Spencer, 2003).
Figure 8: Common features of the analysis phase of qualitative research
It is important to point out that while collecting the questionnaires in a few schools some teachers had not completed filling in the questionnaires by the time of collection. They were personally encouraged to do so and this increased the response rate of the returned questionnaires. As the teachers handed in the completed questionnaires, I would check through them for the ‘fairly obvious’ errors such as omitted answers and failure to follow instructions (Cohen, Manion and Morrison, 2007; Sapsford and Jupp, Ibid p.173). Questions one to four of the questionnaire had closed questions and the remainder (18) were open-ended. The closed questions formed demographic data, which were coded into variables and the values were treated to provide descriptive data while the open-ended questions had to be reduced to a suitable form to enable the analysis (Sapsford and Jupp, 2008; Cohen, Manion and Morrison, 2007). Frequencies and percentages were commonly used in this study. In order to shape the raw data to readily allow for inspection and analysis, familiarisation with the raw data was necessary (Miles and Huberman, 1994; Ritchie and Spencer, 2003; Sapsford and Jupp, 2008). Data were read and various concepts compared, combined, classified and related to other parts and related to previous knowledge (Dey, 2005). The purpose of doing this was to be able to highlight the essential aspects of the phenomena of this study and to summarize the salient features for the purpose of answering the research questions and examining how the responses from participants linked with the entire investigation. All answers were typed out and then listed and re-sorted
162
to identify answers of the same kind, and then grouped and highlighted. They represented answer categories that were of interest in relation to the research objectives. Coding continued on the grounds that the responses were thought to answer the respective questions and/or were of ‘sufficient interest and diversity’ to warrant coding (Sapsford and Jupp, 2008, p166).
In answering the research questions, the participants interpreted and understood their meaning by generating an opinion or reflecting on the past. The open questions from the questionnaires elicited a wide variety of responses (Payne, 2004) which provided background for interpreting answers to the relevant questions. However they took long to administer and responses were difficult to interpret and analyse. Analysis of data depends on study design, number of groups and type of data (Payne, 2004). Researchers can often gauge importance of certain data but coding and interpreting becomes very challenging. Therefore more rigour is vital in interpretation of data with open questions Dey (2005). Although the response rate for the questionnaires was 79.5%, not all questions had a similar response rate and not all questions were answered by the respondents. The response rate depended on the questions and the number of responses given. The questionnaires contained specific contents of interests and phenomena considered worthy of investigations and were administered in different ways. While categorising some of the categories were unclear and overlapping especially where participants were asked to define Special Educational Needs (SEN). While analysing, Implicit and loosely defined classifications of definitions were initially given. The boundaries were not firmly defined and while assigning the categories, the dissimilar aspects were not entirely excluded from others. Possibilities were discounted but not completely excluded. For example, the definitions were classified into different groups in an attempt to sort out and classify different teachers’ responses and to understand what was common or dissimilar in their definitions. The categories were differentiated ensuring that no piece of data fitted into more than one category. They were mutually exclusive. This process was iterative in order to ensure that all data was assigned to at least one category. Where any data did not fit in existing categories more were created
163
to make it exhaustive (Dey, 2005). This was done iteratively. I ensured that the response rate for each question was exhaustively considered.
However, some responses could not be coded because they were incomplete, illegible and some teachers had failed to record answers according to instructions or not answered at all. In such instances, the answers were considered as missing, since they could not be interpreted within the structure of the question or questionnaire as a whole. However, if they were substantial in number and represented new ideas, they were coded separately. Nevertheless, I recognize that maximising responses is a major challenge for any survey and higher response rates add credibility to the results.
Data were refined to a manageable level in order to engage in the data leading to a description and explanation of the social phenomenon under investigation. This entailed looking across the entire range of cases across the data. The real meaning of the original data was retained in the worksheet and saved using a password protected format, forming the basis of key terms, phrases and experiences used by teachers.
‘The methods of generating categories depend on type of data being analysed and the aims, inclinations, knowledge and theoretical sophistication of the researcher' (Dey, 2005, p103).
While generating categories, I relied on the inferences from the data, theoretical issues, intuition and knowledge, taking the varying context of the data into account and relating to the phenomenon under investigation in this study.
Manual analysis was preferred to any computer software in order to facilitate further immersion, scrutiny and navigation of the data while familiarising with the story of the participants and connecting with the data in order to identify related terms and significant relationships across the data. Corbin and Strauss (2008; Green 2007; Dey, 2005) assert that engaging with data contributes to the analysis process. As a qualitative researcher my goal was to understand the teachers’ lived experiences’. I felt that analysing data manually enabled in-depth engagement and understanding of data which was necessary given the open-ended
164
questionnaire and the rich data (text) accumulated from the teachers’ experiences. Like the interviews, an iterative process was used to engage with the data.
Although human judgement was applied throughout the analysis process, it was felt that the combination of other Microsoft Office tools like Microsoft Excel and Microsoft Word together with the iterative process contributed to in-depth engagement. To reiterate, critical friends would check through the steps highlighted and this reduced any effect of bias. On the other hand, this experience contributed to the analysis approach and increased my knowledge which was limited in extensive analysis. Appendix 15 shows how key relevant clusters were mapped from the questionnaires.
The following section discusses the method used to analyse the Kenya Special Needs Education (SNE) Policy Framework (MoE, 2009).