Statistical model on student performance in UTHM by using non-parametric, semi-parametric and parametric survival analysis

• Main Author: Che Hamat, Che Azmeeza
• Format: Thesis
• Language: English; English; English
• Published: 2019
• Subjects: HA29-32 Theory and method of social science statistics
• Online Access: http://eprints.uthm.edu.my/483/1/24p%20CHE%20AZMEEZA%20CHE%20HAMAT.pdf; http://eprints.uthm.edu.my/483/2/CHE%20AZMEEZA%20CHE%20HAMAT%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/483/3/CHE%20AZMEEZA%20CHE%20HAMAT%20WATERMARK.pdf; http://eprints.uthm.edu.my/483/
• Institution: Universiti Tun Hussein Onn Malaysia

Student performance defined as students who are capable to success during their studies. This study explored the use of survival analysis to investigate the performance of Bachelor’s degree students in Universiti Tun Hussein Onn Malaysia (UTHM). The data was collected from the Academic Management Office (AMO), UTHM. The main objective of this study is to estimate the survival rates of students with different entrance qualifications. The study also aim to identify the covariates that dominate the student performance, investigate the performance of Cox model based on the violation of the Proportional Hazard (PH) assumption, compare the model performance by using the survival and Accelerated Failure Time (AFT) models and estimate the time ratio (TR) of student performance in accordance to the selected best model. The survival analysis considered the survival approach such as the Kaplan-Meier (KM) method in the non-parametric method, Cox model in semi-parametric model and survival and AFT models in parametric model. The results revealed that students with STPM-entrance qualification had the highest survival rate compared to Diploma and Matriculation holders. The Cox model in the semi-parametric model identified the GPA, entrance qualification and course as the significant covariates to be included in the study. Faculty covariate was excluded since the p-value insignificant at 90% significance level. The result provided by the Cox model violated the PH assumptions. Then, the performance of the Cox model is less accurate. The invalidation performance of Cox model prompted the need to conduct other parametric survival and AFT models to produce more precise results. As a conclusion, the Log-normal AFT model is the best alternative model to estimate student performance in UTHM and other similar higher educational institution.