Hierarchical linear modeling of student and school effect on academic achievement

Abstract

The purpose of this study was to investigate the factors that affect student performance in primary schools in Kenya. Hierarchical linear modelling (HLM) was used to statistically analyze a data structure where students (level-I) were nested within schools (level-2). Of specific interest was the relationship between student's score (level-I outcome variable) and SES, sex and mother's education of the students (level-I predictor variable) and the student-teacher ratio (level-2 predictor variable). Model testing proceeded in 4 phases: unconstrained (null) model, random intercepts model, means-as-outcome model, and intercepts- and slopes-as-outcomes model. The intercept-only model revealed an ICC of .132. Thus, 13% of the variance in scores is between schools and 87% of the variance in scores is between students within a given school. Because variance existed at both levels of the data structure, predictor variables were' individually added at each level. The random-regression coefficients model was tested using the explanatory variables at student level and all the regression coefficients were statistically significant. Next, the means-as-outcomes model added student-teacher ratio as a level-2 predictor variable. The regression coefficient relating student-teacher ratio to score was positive and statistically significant. Scores were higher in schools with more student-teacher ratio. Finally, the intercepts model and slopes-as-outcomes model were simultaneously tested with all predictor variables tested in the model to test the presence of any interactions between predictor variables. The cross-level interactions indicated that student-teacher ratio is a moderator variable.