Student Level Achievement in Different Public University of Bangladesh: A Comparison Between Multilevel Models and Classical Regression Models

In this study, the main purpose is to consider generalized linear modeling where the regression coefficients are also modeled. The models used here are where slopes and intercepts may vary by group or university. The study time of selected students is a dependent variable and the socio and demographic variables are independent variables. The statistical method of data analysis used for this study is the two-level multilevel modeling. In this analysis, we find the dependent variable is associated with time spends in reading academic books (TSRAB), time spends in reading nonacademic books (TSRNAB), and time spends in reading central library (TSRCL) at the 95% level of significance. Using R, we get three multilevel modeling names as the first model is null or empty model when the intercept is fixed, the second model with random intercepts and fixed slopes and the final or the third model with random intercepts and random slopes. Comparing three models, based on AIC and BIC statistics, we get the final model is preferable to the first two models. The results coefficient of determination (R-square) revealed that the multilevel model is better than the classical regression model.