'Is it possible to obtain the weights of questions in an exam that uses Item Response Theory?
If the results are rather approximations than absolutes, could they be inferred with high precision (> 90%) provided a sufficient amount of data?
What resources/methods could best achieve the desired result, if it is even possible?
Could an IRT model be reverse engineered to find the weights, given the response values and the results?
.
With Item Response Theory, a question's weight could depend on interdependent factors such as:
- Difficulty level: whether a question is answered correctly by many or few people;
- Consistency: the answers provided to other questions, being able to detect guesses.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|