Data analysis

Various data analysis strategies were applied in this study. The analysis process was guided by the research questions and related literature developed earlier in the research proposal. These strategies will be introduced below in relation to the research questions.

4.2.4.1 Qualitative data analysis for Research Question 1

In relation to Research Question 1, qualitative research was conducted and the data was generated through interviews (not used in Paper 1), participant observation and evaluation questionnaires. The data analysis had gone hand in hand with the data collection to build a coherent interpretation. It is emphasised that the process of bringing order, structure and interpretation to a mass of collected data is not neat and does not proceed in a linear fashion (Marshall & Rossman, 2011).

According to Kvale & Brinkmann (2009), the interview analysis can actually begin during the interview. When I was doing the interviews, the students told me what they experienced, felt and did in relation to the topics. In the process, they might discover new meanings in what they had experienced and done in the course. For example, the interviewed students in Paper 2 realised the benefits of doing an activity (student teaching practices at schools) for future personal development when they described it in the interviews. In this way, the interviewees (meaning the students) had provided interpretation or explanation of the data at an early stage.

Transcription was a fundamental aspect of the interview analysis process, during which I made sense of and examined the data. When I was transcribing the complete interviews, I developed a familiarisation with the data and had the chance to mark some important places for later analysis. I did not use computer tools to analyse the transcribed data. By reading and re-reading the data, I gained a greater

and greater sense of the whole series, thus developing a deeper understanding of my research and its ties to theories in the literature.

A meaning condensation analysis was employed in the analysis of the interview data. As Kvale and Brinkmann (2009) suggest, meaning condensation “entails an abridgement of the meanings expressed by the interviewees into shorter formulations” (p. 205). This form can help to analyse the complex interview text by seeking natural meaning units and clarifying the main themes. In the process, the natural meaning units of the text are determined by the researcher after reading through all of the data. Each theme, which dominates a natural meaning unit, is restated by the researcher as simply as possible, according to the researcher’s understanding of the interviewees’ viewpoint in the statements. After that, the researcher interrogates the meaning units in terms of the specific purpose of the study, and finally draws the essential themes of the entire interview into a descriptive statement. During this process, some interviewees were asked for possible corrections or further explanation by email to ensure that the data accurately represented their opinions.

Differing from the analysing of interview data, context analysis (Dörnyei, 2008;

Kvale & Brinkmann, 2009) was used to analyse the data generated by participant observation and qualitative questionnaires for course evaluation. This form often uses categorisation as an essential feature in reducing large data pools (Flick, 2009).

The data pool generated by the two methods in this study was not large due to the small number of participants and the short length of the courses. The purpose of using context analysis was more for showing the key points and categorising the data more systematically. For example, the data from 13 questions in Paper 1 was categorised as motivation, expectations, learning experience and application of knowledge in the future. The categorisation had given a well-structured and simple overview of the data. Observation data analysis followed the same analytical strategy. However, only the results relevant to the discussion were presented in the report.

4.2.4.2 Mixed methods of data analysis for research question 2 In relation to Research Question 2, mixed methods research was conducted, and the data was generated through group interviews, participant observation (not used in Paper 4), and surveys (a post-course survey for Paper 3, pre- and post-course surveys for Paper 4). In the mixed methods design, the data were derived from both quantitative and qualitative perspectives. Thus, two parts were included in the process of mixed methods data analysis: quantitative data analysis and qualitative data analysis. The analytical process applied to the qualitative data (from interviews and participant observation) for Research Question 2 was similar to that used in investigating Research Question 1, and will not be repeated in this section. This

section will focus on the quantitative data analysis in relation to sub-questions 2.1 and 2.2.

According to Creswell and Clark (2011), in quantitative data analysis, the researcher analyses the data based on the type of questions or hypotheses, then uses the appropriate statistical test to address the same. The choice of statistical test is based on the type of question being asked (e.g., a description of trends, a comparison of groups or the relationship among variables), the number of independent and dependent variables, the types of scales being used to measure those variables and whether the variable scores are normally distributed. The quantitative data in both Paper 3 and Paper 4 were analysed using the software SPSS 22.

1) Quantitative data analysis for sub-question 2.1

A post-course survey was used to collect quantitative data in relation to sub-question 2.1: To what degree can tasks make students feel motivated to learn Chinese, and what characteristics do students associate with the motivating tasks?

With the assistance of my co-author Xiaoju Duan, all data were entered into a computer file following three steps: 1) creating the data file in statistical software SPSS; 2) defining the coding frames for the variables; and 3) keying in the data (Dörnyei, 2007). Sub-question 2.1 suggests two types of questions: the description of the motivating degree of each task, and a comparison of the motivating effect among those tasks. Concerning the types of the questions and the number of more than two dependent variables, repeated measures of the general linear model (GLM) were used to analyse the differences among the items. In the process of discussing the results, it was necessary to test the hypothesis arguing that an unfocused task was considered more motivating than a focused task. Thus, a T-test was used to compare the mean values of the two sets of scores, which could provide a statistic to evaluate whether the numerical difference between two means is statistically significant (Hartas, 2010).

2) Quantitative data analysis for sub-question 2.2

A longitudinal, survey-based investigation (consisting of pre- and post-course surveys) was conducted in the study presented in Paper 4, in relation to sub-question 2.2: How do students’ motivational orientations change in a Chinese language and culture course using a TBTL method? All data gathered from the surveys was entered into a file in SPSS in the same method as described above.

Sub-question 2.2 also suggests two types of questions: comparisons among the learner orientations in both pre- and post-course surveys, a comparison between the levels of learner orientations before and after the course, and the relationship among

the orientations and course variables, including task-motivating degree, integration of cultural elements, learner perception of the difficulty of learning Chinese, learner satisfaction and course evaluation. Based on the types of the questions and the number of dependent variables, repeated measures of the general linear model (GLM) were also used to analyse the differences among the orientation in both pre- and post-course surveys. A T-test was used to compare the mean value of each orientation before and after the course, thus finding in what ways and to what extent these orientations changed. Concerning what factors are related to the changes in the given context, the correlations among the orientations and above-mentioned course variables were analysed with the quantitative data from the post-course surveys. It should be noted that the correlations were not directly considered as factual causations, but showed only that some variables were related or correlated (Strand, 2010). In this study, the results of quantitative analysis were explained by the results of the qualitative analysis in the second phase.

In these two studies, the surveys were all filled out in the presence of the teacher, and the students were asked to write their names on them for the research convenience. The group interviews were also conducted by the teacher. However, the teacher did not have any power to influence the students’ course results, and the purpose of the research was clearly explained beforehand. The students were therefore aware that their feedback would be used not to evaluate their teacher but rather to evaluate the tasks they had set, which would be useful for the course’s development. These factors enhanced the validity of the studies.