CHAPTER 3 RESEARCH METHODOLOGY
3.5 Data analysis
Previous section (see Section 3.2.1) in the current study has argued that this study was based on a pragmatic view to investigate Chinese returnees’ reentry experiences at their
home country’s higher education. The nature of this study required mixed methods and strategies. As Creswell (2009) points out, data analysis in mixed methods research is related to the type of research strategies that are chosen for the procedures. In this study,
“mixed methods analysis” (Creswell & Clark, 2011, p. 203), involving both qualitative (description and thematic text analysis) and quantitative (descriptive) strategies, was used to analyse the data. The data from the quantitative analysis were compared with the themes from the qualitative data. That is, the transcriptions of the recorded data from the interviews and the questionnaire surveys from the same participants were analysed and then these two sets of data were compared and merged. Table 3.7 shows the analysis process of the study and the following sections describe it in more detail.
Table 3. 7 Analysis process of the study
Types of methods Types of data analysis Data analysis steps
Mixed concurrent design Merged data concurrently
x Analyse qualitative data (the
primary data set for the study).
x Analyse quantitative data (the
secondary data set for the study).
x Decide how the two data sets
could be compared (e.g. dimension, information).
x Interpret how the primary and
secondary results answer the research questions.
3.5.1 Qualitative data analysis
Creswell and Plano (2011) emphasise that it is important for the researcher to choose the analysis strategy that best suits the study. In the current study, thematic analysis was adopted as the analytical tool because of its flexibility and effectiveness. This approach can be applied to qualitative study, quantitative study or even mixed method study (Miles & Huberman, 1994). Most importantly, it is a method which can be used for identifying, analysing and reporting themes for qualitative data (Braun & Clarke, 2006). It can also minimally .organise and describe the data set in rich detail (Braun & Clarke, 2006). One of the benefits of this analysis method is its flexibility with theoretical freedom, which enables it to be “a flexible and useful research tool”, and “potentially provide a rich and
detailed, yet complex, account of data” (Braun & Clarke, 2006, p. 78). The qualitative data analysis in this study consisted of the analysis of the interviews from the three groups, using a systematic procedure because this “helps to organize [a] large database” (Creswell & Clark, 2011, p. 207). For the current study, the Chinese transcripts from the interviews for the three groups of participants totalled nearly 200 pages. Following
Creswell and Clark’s (2011) and Braun and Clarke’s suggestions, three stages were used to analyse the data: exploring the data, coding the data, and generating categories and themes.
The first stage was to explore the data. The main purpose of this stage was for familiarising the researcher with the data. Though all the data were transcribed by the researcher, it was very important for the researcher to be familiar with all aspects of the data, as doing this provides the bedrock for the rest of the analysis (Braun & Clarke, 2006). This phase included two steps. First, each interview was thoroughly listened before being summarised in memo format (see a sample in Appendix 8) after all the data collection was finished. This enabled the researcher to get a basic idea about each interview and to get a general impression and feeling about the interviews. These memos were short phrases or ideas and included the researcher’s reflections about the interviewees. Creating these memos was an important first step in forming broad categories of information (Creswell & Clark, 2011), which were valuable in the subsequent steps of further clarifying categories and themes in the data analysis. Then all the data were read and reread by the researcher for searching for meanings and patterns. During this step, the researcher began to take notes or use highlighters to underline those parts which the researcher thought were important.
The second stage, coding the data, started when the researcher had read and familiarised with the data, and had generated initial ideas about what was in the data and what was interesting to her. The transcripts were divided into small units such as phrases, sentences or paragraphs according to key words (e.g. academic, salary, new ideas, academic circle, advantages, etc.). As suggested by Creswell and Clark (2011), this process was done by coding directly onto the printed transcript, with code words for text segments recorded in the left margin and broader themes recorded in the right margin. The code words were then sorted into groups of similar meaning and a theme was identified and documented accordingly. During this process, highlighters and coloured pens were used to indicate potential patters or segments of data. Codes were initially identified, and then were matched with data extracts that demonstrated the codes (see Table 3.8 for an example). Each data item was given full and equal attention in the coding process, and interesting aspects in the data set were identified that might form the basis of repeated themes.
Table 3. 8 Codes with data extract
Data extract Coded for
The biggest challenge is to adapt. Since you have come
back to China, you … have to do things according to the
domestic norms here. All these norms, way of doing
research, life styles and rules, you should … re-adapt to them. We cannot change it. Maybe the pain is that I cannot change it, right! Therefore, what you have to do
is … adapt to it.
1. Talked about cultural re- adaptation
2. Attitudes towards re- adjusting
3 .Modify oneself to fit in
Creswell (1997) suggests that to analyse data, all the collected information should first be studied to gain an understanding of the overall data, and then it should be reduced to categories, themes and patterns. Stage three started when all the data had been initially coded. During this stage of data analysis, the researcher’s attention was focused on the broader level of themes and central ideas generated from categories, and all the different codes were sorted into potential themes. Relevant coded data extracts were collated within the identified themes. In this stage, the transcripts were thoroughly examined again, first, to make sure the themes were related to the data set, second, to recode any data within themes that might been missed in earlier coding process. After this stage, five broad initial categorisations were identified:
• reasoning for coming back
• non-returnee colleague and administrators’ views towards returnees
• returnees’ academic adaptation • returnees’ sociocultural adaptation
• returnees’ institutional adaptation.
Cresswell also notes that an important step at this stage is to begin to develop a list of tentative categories and then expand these categories. Thus, the themes from the previous stage were recorded under these categories and grouped according to
interviewees’ common responses with the same implied meaning. This thematic analysis method was very helpful during the analysis process as it contributed to the
understanding of the story line of returnees’ experience and non-returnee colleagues’ and
administrators’ views on returnees.
3.5.2 Quantitative data analysis
As mentioned earlier, in the current research the purpose of the quantitative data was to support the qualitative data, with a set of 5-point Likert scale questions for each group to investigate their respective perceptions of and attitudes towards returnees. The survey questions aimed to find out how much the participants agreed or disagreed with the statements and all the questions were closed questions. The essential characteristics of the data were described in quantitative forms of frequency distribution and graphical displays (Johnson & Christensen, 2012). The analysis technologies used were
Microsoft Excel and SPSS, which were used together to analyse interviewees’
demographic information, frequency of variables, range, mean and median. As the total number of questionnaires was relatively small (44 in total), the figures of the data were manually entered by the researcher into Excel worksheet and SPSS sheets. Excel technologies were used to display the percentages of the data, whereas SPSS was found more appropriate to explore such information as frequency, mean and median of the data. During the analysis process, five items in the questionnaire surveys for non-returnees (items 14, 15, 16, 18 and 19) and items (16, 17, 19, 20 and 21) for administrators were reworded, according to SPSS requirements, because they had been negatively worded. Creswell and Plano (2011) suggest that once the mixed methods data analysis is complete, mixed methods interpretation should then be used to examine the quantitative results and qualitative findings, as data comparison and integration is important for mixed research studies (Johnson & Christensen, 2012). For the current study, data transformation, as suggested by Onwuegbuzie and Teddlie (2003), was used to analyse the data, with the themes or factors created from the quantitative data compared with the themes and categories from the qualitative data. A matrix was created to display the combined information from both databases. A variety of tabular and graphical forms, as well as charts, were used to display qualitative as well as quantitative data. When displaying qualitative data, tables were applied to display the data, with columns indicating numbers
of participants, content, and the researcher’s comments, Quantitative data were mainly used to support theses and categories from qualitative data because it was in a supportive
role in the study as mentioned earlier. All the findings were reported in the next two chapters (Chapter 4 and chapter 5).