Data mine good questions
- Project: Data mine WW database for good/bad questions. Write software that allows one to determine which WeBWorK questions are hard, which ones are defective? which ones are effective?
- The basic plan would be to data mine the databases existing at each site to catalog questions that have an unusual number of wrong attempts before the question is answered correctly. This requires software to do the preliminary extraction and then software to make it fairly easy for humans to analyze the questions and determine what causes the large number of wrong attempts. Bad question wording? just hard? used in inappropriate course? How to determine effectiveness is even more nebulous -- although that is what would most like to know. Another statistic might be which questions are most widely used? A flexible tool that allowed asking many related questions would be most useful.
- Add other statistical questions here:
- Preliminary plan of attack: Data mine WW database details
- Related project: User interface for question statistics. Develop user interface software that mapped collection of statitical metadata about questions described above onto the NationalProblemLibrary and Library browser so that instructors would have this information easily available as they chose questions for their classes. This would probably involve perl programming to interoperate properly with the existing WW course management interface.