Application of log-linear models in analysis of students' mathematics anxiety

Abstract

A log-linear modeling for three-dimensional contingency tables is used with categorical variables. In this study, hierarchical log-linear models are used to fit the observed frequencies and to determine the suitable model for the expected frequencies of pre-university students? mathematics anxiety data. For this reason, 546 students were selected to complete mathematics anxiety scale questionnaire. Data on gender, course, grade and level of mathematics anxiety were examined for presence of association using log-linear models. Estimating log-linear model parameters is carried out using Maximum Likelihood method, where the Newton-Raphson iteration method is used to find the numerical estimates of the parameters. Selection of the best model is conducted using deviance of the models. The significance of the model is determined using Goodness of Fit Test. We also determine the odds ratio between gender, course, Sijil Pelajaran Malaysia (SPM) Additional Mathematics grade and anxiety level to establish the risk among the groups. The final model shows that gender, course and SPM Additional Mathematics score play a role in determining the students' level of mathematics anxiety.