I'm trying to get the KR20 value for testing and have run into a snag and think I understand why but was hoping someone could confirm or correct me.
All references and examples I've seen show that all persons are given the same set of questions. In our data set, we have a pool of 77 questions (for this specific test) from which 10 questions will be chosen/presented to the students. My problem, I think, comes from the fact that the sum of the squares of the (scores - mean) is summing up the results from all 77 questions and not just the 10.
Is there a variation or something that can handle the issue I've described or is there another formula that would apply in this instance instead of the KR20 formula?