Research Information System
Van Yüzüncü Yıl Üniversitesi Eğitim Fakültesi Dergisi, vol.21, no.3, pp.840-868, 2024 (Peer-Reviewed Journal)
International large-scale assessments have a key role in improving educational, economical, and political systems. By using the data of these assessments, countries can draw conclusions about the status of educational systems. Studies and reports generally tend to choose variables available in data set to model the relationships among the variables. In this study, we aimed to introduce a variable selection method to analyze large-scale assessments to be able to decide which variables might be included in modelling country data. We used the entire data set of Türkiye PISA 2015 through elastic net regression to decide which variables should be selected for further analysis. We also provided a summary of the available studies based on Türkiye PISA 2015 data and compared the results. Based on the series of analyses, this study revealed that test anxiety, environmental awareness, interest in broad topics in science, playing video games after school, mathematics literacy, reading literacy, and collaborative problem-solving skills were the explanatory variables contributed most to the degree of scientific literacy of students. This study has a potential to provide an example of shrinkage methods applied in educational context and offer another standpoint for providing a rationale to select which variables can be included.