Multilevel linear models in SPSS in educational research

Abstract

In this article, we analyse data that are grouped or nested in larger units, as is often the case in education. To do this, we use the database that the PISA test (OECD) provides, in which students (level 1) are grouped in centres (level 2). The aim is to illustrate the analysis that best suits this type of hierarchy structured data following multilevel linear models (MLM). MLM are determined using the SPSS program and considering the groupings. Thus, we analyse how the independent variables of level 1 (personal and familiar) and those of level 2 (all related to the centre) influence results related to students' performance in the PISA 2015 test (dependent variable). In the example, we demonstrate not only that the variables of both levels 1 and 2 had a considerable impact on students’ results, but also that the analysis through MLM considers inter-level effects, in this case grouping in centres, in the result.