Project topics and materials
Get Complete Project Material File(s) Now! »
RELIABILITY ANALYSES
Once the items were selected according to factor analysis, it was important to confirm the reliability of a scale consisting of the items extracted. Reliability analysis was carried out for these items. It is said that a factor is reliable when it has an alpha coefficient of at least 0.65 (DeVillis, 1991). Nonetheless, because the current study is exploratory, 0.5 is acceptable (Howie, 2002). Besides alpha coefficients, „Corrected Item-Total Correlation‟ and „Alpha if item deleted‟ were examined as a means of selection. At student level in Korea, factors extracted from the factor analysis conducted in the previous section were examined to see if they had internal consistency to build a sound construct. Factors or scales that consist of two or less items were not carried out for reliability analysis. In addition, reliability analyses were carried out on dichotomous format (yes/no) questions. This was the case with „home possession‟ and „safety in school‟. „Home possession‟ consists of 16 sub-items, and although the factor analysis supports a factor „home possession‟ conceptually, it was excluded in the factor analyses as it is not Likert scale, but “yes/no” format. Even after examining reliability analysis, „home possession‟ was excluded for further study since the reliability coefficient was still too low (see Appendix G). Regarding learning activities, „STS learning‟ only has three items remaining, as two items were deleted due to low communalities (in Section7.2) and low Corrected Item-Total Correlation values respectively. Nonetheless, it still shows a meaningful alpha coefficient, 0.619. On the other hand, „play after school‟ also has three items because one item, „play with friend‟, was deleted due to low Corrected Item-Total Correlation value (0.281) and alpha value of 0.636 (see Appendix G). As seen in Table 7.39 (above), factors examined satisfied the criterion value, which is Cronbach alpha=0.5, even though most of the factors consist of a few items that remained.
CORRELATION ANALYSES
As the last stage of preliminary analyses, correlation analyses were carried out for the scales or factors identified up to this point. Items consisting of a question were previously examined in terms of factor analysis and reliability analysis. Once it was confirmed that the items underlie one construct and have internal consistency, they were put together to comprise one scale. Next, variable names and labels were assigned for further analysis, and these scales were reexamined by means of correlation analysis. In addition, single-item factors considered important to student achievement from a conceptual point of view were investigated in terms of correlation analysis.
First, the bivariate correlations were examined between the scales or singleitem factors and science achievement (see Appendix I and J). Next, the intercorrelations were analyzed between the scales or single-item factors. The intercorrelations were explored to identify whether mulitcollinearity, which is an assumption for regression analysis, was present. The bivariate Pearson product-moment correlation coefficient γ was calculated. The scales or singleitem factors that have a correlation coefficient of an absolute value above 0.15 are described and discussed in the following sections. This cut-off point for exploration, not for inclusion for further analyses, was chosen to preliminarily identify possible relationships with science achievement as it was used in some exploratory research previously conducted (Bos, 2002; Howie, 2002). The variance explained also has to be considered, to ascertain how much variance is shared. The variance explained is calculated by squaring and multiplying γ value by 100 to make a change into percentage of variance (Cohen et al., 2007).
STUDENT LEVEL
The results of correlation analyses were explored, starting from the student level of Korea to South Africa. In addition, comparison between the two countries was made, corresponding to factors examined, and helping to answer the first research question.
CHAPTER 1 INTRODUCING THE STUDY
1.1 Introduction
1.2 Education in Korea
1.3 Education in South Africa
1.4 Problem statement
1.5 Rationale for the study
1.6 Aims of the study
1.7 Research questions
1.8 Structure of the dissertation
1.9 Conclusion
CHAPTER 2 BACKGROUND TO THE TIMSS 2003 STUDY
2.1 Introduction
2.2 Design issues regarding TIMSS
2.3 Research design
2.4 Instruments
2.5 Data transformation
2.6 Data quality
2.7 Conclusion
CHAPTER 3 RESEARCH ON FACTORS INFLUENCING STUDENT PERFORMANCE
3.1 Introduction
3.2 School effectiveness research
3.3 Factors related to science achievement
3.4 Conclusion
CHAPTER 4 CONCEPTUAL FRAMEWORK FOR THE STUDY
4.1 Introduction
4.2 Creemers‟ model
4.3 Scheerens‟ model
4.4 Shavelson, McDonnell, and Oakes‟ model
4.5 Conceptual framework for the research
4.6 Conclusion
CHAPTER 5 RESEARCH DESIGN AND METHODOLOGY
5.1 Introduction
5.2 Secondary analyses of data
5.3 Discussion of research questions
5.4 Sample
5.5 Data collection
5.6 Instruments
5.7 Data analysis
5.8 Correlation analysis
5.9 Multilevel analysis
5.10 Methodological norms
5.11 Ethical considerations
5.12 Conclusion
CHAPTER 6 DESCRIPTIVE ANALYSES
6.1 Introduction
6.2 TIMSS science achievement scores in Korea and South Africa
6.3 Exploring the data sets
6.4 Conclusion
CHAPTER 7 RESULTS OF FACTOR, RELIABILTY AND CORRELATION ANALYSES
7.1 Introduction
7.2 Factor analyses
7.3 Reliability analyses
7.4 Correlation analyses
7.5 Selection of variables
7.6 Conclusion
CHAPTER 8 RESULTS OF MULTILEVEL ANALYSES
8.1 Introduction
8.2 Preparation of the data
8.3 The results of the multilevel analyses
8.4 Conclusion
CHAPTER 9 CONCLUSIONS AND RECOMMENDATIONS
9.1 Introduction
9.2 Summary and the research questions
9.3 Discussion and reflection
9.4 Recommendations
9.5 Conclusion
REFERENCES
