Computing the Number of Attempts Needed On An Assignment to Improve Course Design

Abstract

With recent surges in online learning, much of which can be attributed to COVID, homework, and quizzes are easily able to be graded and retaken. My goal was to use course data to figure out the optimal number of attempts a student should get for a course assignment to see the greatest improvement in the class. Multiple studies such as Mackenzie (2015) and Kortemyer (2019) show that having more attempts allows for increased retention of the topic. They both also found, however, that having unlimited retakes hampers student retention. Finding the threshold number of attempts necessary to see the largest improvements in retention is important to making online learning more effective. To find this I initially used a polynomial regression line with the independent variable being the number of attempts a student took and the dependent variable is the average rate of improvement of their attempts. This was calculated by adding the difference in the grades on each attempt and then dividing it by the total number of attempts. When I graphed these values there was a weak correlation. To create a stronger correlation, I conducted a multivariable polynomial regression with the two independent variables being the number of attempts and the average duration of each attempt (to capture students who were speeding through the questions). This provided a much stronger correlation to the improvement rate. I then proceeded to graph the best fit line I created and found that there were moderate increases in performance up to 3 attempts.