Data analysis as a decision support tool in Greek education

Abstract

This dissertation highlights the potential offered by the analysis of educational data in primary and secondary education. Starting from a common issue in educational research— the academic achievement of students—it emphasizes the potential for objective evaluation and decision support provided by the use of data in centralized educational systems. Central to this is the concept of equal opportunity education, which runs throughout the thesis. Since the industrial revolution, education has been seen as a tool for social mobility and equal opportunities. States introduced compulsory public education and financed it through taxation. A main objective remains equality of educational opportunities for students through public provision and state control. But research has highlighted that long-standing social inequalities are also reproduced through education systems, leading to proposals for more inclusive education systems and educational interventions. The evaluation of the effectiveness of education systems and the expression of views in the public debate often reflect personal perceptions, which are not based on objective evidence. Recently in Greece, systematic collection of educational data has become possible with the introduction of a MIS for primary and secondary education, but the potential for knowledge extraction from it has not yet been exploited. By examining student achievement in Greece, this thesis highlights the potential benefit of using data analysis in evidence-based decision-making and drawing objective conclusions. The use of these tools provides critical knowledge for decision-making by policymakers and educational administrators. The majority of educational data analysis research on student achievement focuses on higher education and online learning. Additionally, studies often use small sample sizes, which may limit their generalizability. Longitudinal analyses, which could reveal the long-term effects of educational interventions, are scarce in the literature. This thesis analyzed the entire student population of the country, both statically and longitudinally, drawing objective conclusions on dimensions of the education system as a whole as well as individual educational interventions. In addition, it broadens the research scope to tional levels with different characteristics from higher education, which have a significant impact on students and society. The research questions of the thesis are related to students’ academic achievement. The first research question focuses on the objective detection of different levels of student achievement and forms the basis for further analyses. The second research question examined the stability of the identified achievement levels over time. The third research question examined the function of the school as an equal opportunity institution through the impact of demographic (non-academic) characteristics, such as gender, guardian occupation, and region of residence, on academic achievement. The fourth research question examined the potential for objective evaluation of a specific educational intervention, that of remedial teaching, in the light of equal opportunities for students. Finally, the fifth and last research question examined the predictive power of GPA in estimating future achievement against alternative, weighted metrics with different weights of courses. To meet our research approach, we requested demographic and academic data of the country’s students from the Ministry of Education. We obtained data of the entire student population, from 5 of primary school to grade 3 of Junior High School. The data were: a) Grades in all subjects b) The class of each student c) The overall Grade Point Average d) The students’ absences e) The gender of the students (f) The profession of the guardian (g) The education directorate to which each pupil belonged. The school years for which we obtained data were from 2016-17 to 2018-19. In this thesis, unsupervised learning was used to assess student achievement to reduce researcher intervention. The algorithm added each student’s achievement level to the dataset and ranked them by achievement level. This variable was used to answer remaining research questions, such as student achievement differences by gender, region, and guardian occupation. Finally, a longitudinal analysis of achievement levels from grade to grade examined student achievement stability. The thesis also showed that data analysis can yield meaningful conclusions from educational data, even if it was not collected for research. It used national student data for the first time to categorize them into four mathematically calculated academic achievement categories. The longitudinal study of student achievement found stability over time, with the highest and lowest performing students showing strong stability. It was also found that the level of student achievement was influenced by non-academic factors such as gender, region of residence, and the guardian’s occupation. The nonindependence of achievement on non-academic characteristics provides clear evidence in favor of the argument that the education system does not function as a system of equal opportunities for students. The research further found that remedial teaching had short- and long-term effects on the improvement of students overall, but the improvement differed by the profession of the guardian, favoring more privileged students. This demonstrates the opposite effect of remedial teaching from its objective, which is to enhance equal opportunities for pupils who have socially limited opportunities. In terms of contributions, this is the first research effort using data analysis at the country level. Our study results relate to all students in the country without the need for statistical inference. It was found that the use of educational data, which already exists in the databases of the Ministry of Education even if not collected for a specific research purpose, can lead to informed opinions on the functioning of the educational system. This allows the ministry’s services to engage in in-depth analysis of education data to extract new knowledge that is currently ”hidden” in the large volume of MIS data. An approach has been developed, that of objective identification of achievement levels through clustering, which can be used in other researches on student achievement, without the need to study distributions of student grades. It was found that there is an objective way of dividing achievement levels and characterizing student achievement. From this procedure, specific and numerically stable achievement levels emerge, highlighting corresponding stability in the set of factors affecting achievement while posing challenges to educational policy. As the data do not support the achievement of the target, the pursuit of an equal opportunities school should continue. The differentiation in performance between students from different social and economic backgrounds shows that further efforts are needed in order for the school to function as a tool for social mobility and equal opportunities through pupil achievement. For the first time, using total data, the overachievement of girls compared to boys has been confirmed. Similar studies have used data from student competitions, such as PISA, with a limited number of students and subjects tested or small samples. The thesis confirmed the findings for the first time at the country level, without the need to induce the results. Overall, through the evidence-based assessment of dimensions of the education system and educational interventions, it became clear that the analysis of our country’s educational data provides enormous potential for informing decision-making and evaluating the outcomes of educational policies. Thus, the need to integrate information system data into the decision-making process is emphasized, as well as the importance of promoting data-based decision-making in education.