Question: I was looking at the item analysis for a test I administered and I noticed that most of the point-biserial correlations were zero or close to zero. Is this a problem?
Answer: The point-biserial correlation is a measure of item discrimination: how well a test item discriminates between good performers and poor performers. A point-biserial correlation of zero, or close to zero, tells us that the item doesn’t discriminate at all. Why? In a mastery test in which the passing score is high (85-90%) and everyone, or almost everyone, passes it’s not surprising that most items don’t discriminate — all the test takers will have answered most items correctly. There are no poor performers. Is this bad? That depends on your point of view — but you are losing valuable information about your training and your test takers by giving “easy” tests that everyone passes.