A longitudinal study of student performance in English using repeated measures, multilevel and logistic regression models

Abstract

This paper presents three statistical models that analyze longitudinal data on student performance in English. A random sample, comprising male and female students who attend either a state or a private school, was selected to investigate gender and school bias in this subject. The English annual marks attained by each student were recorded during the last three years in primary schools. In the first approach, we present a repeated measures analysis of variance that captures the correlation between the repeated measures. Several tests are carried out to check for within subjects and between subjects effects; equality of covariance matrices and sphericity. In the second approach, we fit a two-level random coefficient model to examine the effect of time on student performance in English. This model allows the student-specific coefficients describing individual trajectories to vary randomly. In the third approach, we fit a Logistic regression model to estimate the probability of passing the Eleven-Plus examination that students sit for when they terminate Primary education.