Modeling Student Performance in Mathematics Using Binary Logistic Regression at Selected Secondary Schools A Case Study of Mtwara Municipality and Ilemela District

Abstract

This study investigated the performance of secondary school students in Mathematics at the Selected Secondary Schools in Mtwara Municipality and Ilemela District by Absenteeism, Conduct, Type of School and Gender as explanatory Factors. The data used in the study was collected from documented records of 250 form three students with 1:1 gender ratio – 50 students from each of the five selected secondary schools in the academic year 2011/2012. The sample was considered appropriate as they had covered more than half of the Mathematics syllabus in Ordinary Secondary Schools. Binary logistic regression was used to model a binary variable ‘performance’ (fail, pass) against a systematic component of linear combination predictors (absenteeism, conduct, type of school and gender). The model fitted for the log-odds in favour of poor performance is . The essence of this study is to provide student performance analysis method (Binary Logistic Regression) not commonly used in Tanzania. Findings show that two out of four explanatory factors used in the study (absenteeism and misconduct) significantly predict student performance in Mathematics based on binary logistic regression fitted. Absenteeism and misconduct predict the log-odds of poor performance by multiplicative effect of 1.414 and 3.137 respectively. Future work is recommended to focus on analysis using other Generalized Linear Models (GLM) as well considering other locations with more/other variables affecting performance of students in mathematics.

Keywords: Binary Logistic Modeling, Misconduct, Performance.