Teacher's AI aide: Code for analyzing multiple choice test results Link to Website
This repo has python scripts for data mining multiple choice test results. It has been designed to work on top of existing flubaroo output. It is hoped that additional analytics can help teachers design better multiple choice exams and learn something more about their students. This project is under sporadic development.
TODO
Here are instructions for teachers to get the code to run on their own text outputs.
- Windows TODO
- Mac TODO
- Distribution of scores TODO
- Percentile distribution TODO
- Question Difficulty TODO
- What percent of students got this questions wrong?
- If students were randomly guessing, how likely would the observed answer be?
- Misconception Analysis TODO
- Is a student's overall performance predictive of what they answered on a question?
- Question Discrimination TODO
- How predictive is a question of the student's overall performance?
- Frequent Patterns TODO
- Is there an association rule between questions? Ex) answering c on question B means students answer a on question 2 with very high confidence.
- Student Clusters TODO
- Student's who have similar understandings will likely have similar answers on the test. We can try to automatically group these students.
- Visualization TODO
- Similarities among students can be visualized in a 2D Plot via a class of methods known as manifold embedding and dimensionality reduction
- Co-Clustering TODO
todo = 'in progress'
TODO