Researcher(s)
- Jeffrey Woulfe, Physics Education, University of Delaware
Faculty Mentor(s)
- Adebanjo Oriade, Physics, University of Delaware
Abstract
As a future teacher, I was very curious as to how I could structure my tests in the future to benefit my students’ performance. I used the test scores from previous years of the introductory physical science course, SCEN101, taught by my research advisor Dr. Adebanjo Oriade to determine if there is a correlation between question order and overall student exam performance. Dr. Oriade gave 5 separate versions of a 70 multiple choice question exam, reordering the questions on each form of the exam. The majority of my task was creating a working item analysis Python code to find the difficulty index and discrimination index for each question on each exam. Difficulty index judges how easy a question is by dividing the total number of students that answered the question correctly, by the total number of students that attempted the question. The discrimination index looks at all questions in the exam and judges how well high-scoring students do on a particular question. Having only taken an introductory computer science class, this proved to be the most difficult and time-consuming part of my research. I had to make sure that the code was dynamic to accommodate any number of questions or students. Once my item analysis code was finished, I took the 70-question exam for myself to see what I thought the most difficult questions were. Then using the discrimination and difficulty indices I can judge the differences in the performance of students on the same questions whether they were placed earlier or later on the test. This is also a good reason for students to implement the strategy of skipping more difficult questions and saving them for later, as it is a good test taking strategy so they do not feel discouraged for the rest of the exam.