Student Usage of Small Auto-graded MATLAB Coding Exercises

Abstract

Instructors are increasingly using small auto-graded coding exercises with immediate feedback
to help students learn the MATLAB programming language. Such exercises may require
students to write 3 – 10 lines of code. We analyzed student usage of 38 instances of MATLAB
coding exercise instances across 1,435 students from seven courses at different universities to
determine how students are using the automated MATLAB assessment tool. When instructors
suggested completing the exercise (not necessarily requiring or awarding points), we found that
student completion rates were on average 83%, with an average per exercise ranging from 64%
to 95%. We found that students spent 7.8 minutes on average, matching the 3–10 minutes
expected by the exercise authors. We found that students made 4.5 attempts on average per
exercise. For some harder exercises, the averages were higher at 12.5 attempts on average and
10.4 minutes on average, suggesting that students were indeed putting forth good effort. Further,
we analyzed the students’ wrong submissions of exercises that had a high average number of
tries. We identified common mistakes by students and shared our best practices for authoring
coding exercises.