Informatics in Education

We focus on two types of centralised national examinations (the 10th grade tests and Matura examination) that are being carried out in Lithuania for two decades. The aim of the paper is to analyse assessments of mathematics for the entire Lithuanian secondary school population that have no sampling errors while considering the factors of location, school ownership and gender as important indicators when judging about educational effectiveness in terms of quality and equity. We analyse the results of the 10th grade tests for the 2011–2015 period and the results of the same cohorts participating in the Matura examination. We observe that the distribution of the assessments of both exams is asymmetric with a positive skew. The median often is below the middle of the grade scale indicating poor performance or mismatch between knowledge and examination tasks. There are limited differences in assessments with respect to gender and school location, although we detect a tendency to have better mathematics achievement in private schools. The conclusions drawn from national assessment data is somewhat different from international data thus one cannot neglect national information for the development of educational policy. The variables analysed in the analysis has limited predictive power for achievements in mathematics and further analysis is called-for.

Many countries have focused on the improvement of education system performance. Small number of studies consider system of a country as unit of assessment where indicators represent all levels of education system. In the paper, we propose the methodology for the performance analysis of education systems as a whole hybridizing Data Envelopment Analysis and Principal Component Analysis. Its applicability is illustrated by the analysis of the data collected for 29 European countries. In the analysis we used publicly available data from EUROSTAT and OECD which European Commission uses for the performance monitoring of education in European Union. No prior assumptions were made or expert judgements included. We demonstrated good performance of the method on limited data set. The proposed methodology of hybrid Data Envelopment Analysis and Principal Component Analysis allows researchers analyse education systems quantitatively. The recommendations for improvements and assessment of real world education systems should be based on the analysis of a sufficiently large data set comprehensively representing the considered education systems.