An investigation on students' academic performance via Lasso-type regression analysis methods

Abstract

This chapter aims (i) to predict the students' academic performance using a large number of educational data in both Turkey and Singapore and (ii) to answer which educational inputs are really important for the prediction of the educational performance using machine learning methods. Nowadays, machine learning methods have adopted slowly in applied economics and econometrics, even if their scope and purpose in are different from econometrics, especially when it comes to causal inference. Fortunately, recents surveys explore how machine learning can be used in estimation of structural models. This chapter focus on some of these recent methods called Rigorous penalization approach for Lasso and Squared Lasso and Post-Lasso introduced by Belloni, Chernozhukov, Hansen and their collaborators. The contents and results of the chapter can be useful for the researchers who use recent adopted machine learning methods in applied economics and econometrics. To use these new methods can be used for prediction and model estimation, especially if the number of variables are large and/or they exceed the sample size. In the study, the findings of these methods can help to shed light on the existing theories and literature on education. © Peter Lang AG 2019.