Adaptive Testing

We are using the Rasch Model. There is 1 parameter for questions (items), which is the difficulty. This difficulty is supposed to be the number of std deviations away from the mean the question's "hardness" is. So it spans from -3 to 3 since that is close enough to 100% of all questions.

• Questions can be skipped a limited number of times before requiring an answer
• A certain number of questions must be shown since a question was last skipped for it to be shown to the user next.
• Questions can be seen a limited number of times (configurable to 1 as well)
• Questions are chosen from a "window" of difficulty around the user's estimated score. If there are no suitable questions within this window, the user is prompted to increase the size of the window. If the user chooses to, it will increase until at least a single question becomes available
• If the user's variance falls below a certain threshold they are prompted ONCE to stop studying this subtopic if they are happy with their score

I don't think I missed anything