3.6 Data Collection and Analysis Procedures
3.6.2 Data Analysis Procedures
The questionnaire was administered two weeks prior to the end of the academic semester. The numeric Likert-type data collected underwent several types of analysis using the Statistical Package for Social Sciences (SPSS). Firstly, descriptive statistics were used to describe and present the dataset (Cohen et al., 2007). Thus, thefrequencies and percentages were computed to describe the students’ distribution across gender, age, language proficiency and technology skills. The mean scores and range of responses were also calculated to determine the average score and the distance between the highest and lowest scores in respondents’ answers respectively. Moreover, the variance and standard deviation (SD) were computed to determine
83
how far the respondents’ scores were from the mean scores and how dispersed they were. These approaches helped to obtain a general idea of students’ responses.
Inferential data analysis tests were also performed. According to Cohen et al. (2007) and Larson-Hall (2010), inferential statistics are useful to make inferences and predictions based on the collected data. This type of analysis was performed first to identify whether there was variance in students’ responses about their engagement in terms of gender, age, linguistic proficiency and technology skills. It was also used to check whether the variances were statistically significant, i.e. they were not attributed to mere chance, and that a real correlation between the dependent variable (i.e. student perceptions of their engagement) and independent variables (gender, age, linguistic proficiency and technology skills) actually existed (Cohen et al., 2007).
One of the inferential tests performed in the study was theMann-Whitney U test. This test is used to compare differences between two independent groups. It enables us to see whether there are differences between male and female respondents on a rating scale (Cohen et al., 2007). There are four necessary conditions to perform this test: (1) the dependent variable is measured at the ordinal level (e.g. Likert scale); (2) the independent variable consists of two categorical, independent groups (e.g. male/female); (3) no relationship exists between the observations within each group and between the two groups (i.e. no participant can be in more than one group); and (4) the two independent variables are not normally distributed (Laerd Statistics, n.d.). This test was useful to identify whether there was a statistically significant difference between the male and female study participants’ responses.
Another inferential test that was performed in the study was the Kruskal-Wallis test. According to Cohen et al. (2007), this test enables the analysis of variance for three or more
84
independent samples on an ordinal variable. Four major assumptions must be met to perform this test: (1) the dependent variable must be measured at the ordinal level; (2) the independent variable should consist of at least three or more categorical, independent groups; (3) observations in each group or between the groups must be independent; and (4) the distributions of each group’s scores have the same shape to be able to compare the medians of the dependent variable. The Kruskal-Wallis test was employed in this study to check variances in students’ scores based on the age group they belonged to, their linguistic proficiency, and their technology skills. Although de Winter and Dodou (2010) argue that parametric and non-parametric procedures are equally useful to analyse Likert-type data, Frost (2016) explains that these tests are particularly useful when the sample sizes are unequal, which is the case in the current study. The Kruskal- Wallis test did not identify clearly where the difference resided, however. Field (2009) and Larson-Hall (2010) recommend performing several Mann-Whitney U tests in order to overcome this problem, which was the procedure followed in this study.
The recorded qualitative interview data was translated by a professional translator and transcribed immediately after the interviews were concluded. The transcriptions were then cross- checked by the participants and later fed into Nvivo qualitative data analysis software to perform the necessary analysis activities to answer the research questions. Three main processes of qualitative data analysis activity were followed: data reduction, data display, and drawing and verifying conclusions (Miles & Huberman, 1994). According to Miles and Huberman (1994), Punch (2005) and Robson (2002), these three data analysis streams are non-linear, take place concurrently, and are cyclical in nature (see Figure 3.1).
85
Figure 3.1 Components of Data Analysis: Interactive Model (Source: Miles & Huberman, 1994, p. 12)
Two main processes were followed during the data reduction. Firstly, a general sense of the data was obtained through an initial descriptive coding activity, where important data relevant to the research topic and questions was highlighted and labelled while considering the pre-specified coding scheme the study required (Dörnyei, 2007; Punch, 2005). A second level of more inferential coding was then applied to the data using information from the descriptive analysis to identify patterns and recurrent themes.
Although the study necessitated the development of a pre-specified thematic framework and coding scheme to answer the first research question, a general framework involving inductive analytical procedures was adopted simultaneously in order not to miss any other codes and themes the data itself suggested (Adu, 2013; Punch, 2005) (see Appendix VI). Relationships between different themes were then identified, enabling comparisons to be made and conclusions to be drawn to answer the research questions. The conclusions were also verified by checking
86
their representativeness and cross-validating them through triangulation (Miles & Huberman, 1994; Punch, 2005).
This practice-based study is centred on the integration of flipping in EFL writing instruction. A review of the procedures followed in the flipped writing class is provided in the next section.