Statistical Modeling and Analysis of Online Examinations in Covid-19

Abstract

In this paper, the authors build a model that predicts the grade point from a collection of independent variables including student characteristics and exam marks achieved in four marketing management courses using data from four courses. The data from four courses in marketing management, for an off-line class in a master’s in business management program taught in 2019 and the same set of four courses in marketing management taught to an on-line class in 2020 was taken. Although the number of students enrolled was almost equal, and the courses, despite being offered a year apart, were nearly comparable in structure and content, the teaching and assessment for 2019 was conducted offline, whereas it was conducted online in 2020. In the set of four courses offered in 2019 using the offline mode, the exam was proctored and offline but for the same courses offered in 2020, the final exam was also proctored but online. The authors predicted that if exams were taken without any misconduct, the prediction model would have the same explanatory power for all exams, and that if there was malpractice, the explanatory power would be lower. Their findings show that the two datasets are similar; there are variations in the independent variables that statistically and significantly predicted the Grade Point Average (GPA). The R-squared statistic suggests that the model for prediction is strong. Hence there is reason to believe that malpractice was taking place when the examinations were online, in spite of it being proctored. The goal of this paper is to provide teachers with practical ideas for administering proctored tests in their online courses.