Applying explanatory analysis in education using different regression methods

Abstract

Measuring how a college is successful relies heavily on its outcome (i.e., students of the institution). After spending a few years in a college, students will join organizations where they can apply knowledge and skills acquired during the study-life. Therefore, it is vital to ensure that students are well treated, and to achieve that we need to understand how to improve the education environment. To improve an education environment, we need to learn that from factors that impact on success or failure. Data mining studies in education can be descriptive, predictive, and explanatory (i.e., diagnostic). Although Predictive models can tell what would very likely to happen when certain factors are present, they cannot tell how these were occurred. Therefore, explanatory models can explain how underlying factors are exist and can quantify their level existence which will lead to improving education practice in general. Underlying factors include independent variables (e.g., gender, age, disability) and the interaction between these variables. In this paper, we define potential methods that can help to provide explanatory studies using educational data. Also, we define machine learning algorithms (i.e., regression tools) that can be used for this type of study including preprocessing the data, test of multicollinearity of the specified model, interactions involvement, and model validation. In addition, we presented a case study using synthetic data to explain how this method is implemented. In the case study, we explained variables and interactions contributed to students scores. Also, we reported performance measures used for the linear outcome.