Predicting Students’ Final Exam Grades Using Cumulative Odds Model

The retaining and performance of students in science and mathematics is increasingly recognized by higher education institutions in Malaysia. The study of the factors that determine student success is the first step in improving their performance and preventing student disaffection in their studies. This study offers a comprehensive approach to developing predictive models of student final exams through the ordinal regression model with academic process factor. The proposed model is the least used model in previous studies despite the availability of ordinal categorical data as the response variable. Predictive performance was evaluated using measures of goodness of fit test, bias, and RMSE. The model fitting to real data sets has shown that academic process factor plays an important role in the performance of the final exam for a mathematical statistics course. The cumulative odds equation of the model that was generated can be used by the institution of study chosen to predict the performance of the final exam of their students for the mathematical statistics course. Simulation analysis results show that the cumulative odds model (COM) is able to manage big data through software R. The results from the simulations also show that the model parameter estimates become more consistent when sample sizes are increased. The choice of R software for the process of fitting is accurate because the assumptions of the ordinal regression model can be easily checked using the latest methods.