Exam Analytics: Point Biserial & Discrimination Index

Exam analytics: What is point biserial? Some people call this the correlation coefficient. What is discrimination index? The point biserial is a measure of how success on the exam correlates to success on a specific item. It ranges from 0 to 1 with higher numbers showing a stronger correlation. So, higher is better. If this was a mastery item on a concept we’ve taught so many times we expect everyone to get it correct, you might see a 0 here, because they should get the item correct whether or not they did well on the exam. The discrimination index is a measure of how success on a specific item correlates to success on the exam. It’s sort of the reverse of the point biserial. It also ranges from 0 to 1 with higher numbers showing a stronger correlation. So, again, higher is better. Keep in mind that those mastery items may show zeros here because they’re intended to confirm knowledge rather than to discriminate between different levels of performance. Looking back at this summary, the second item has a very low point biserial and discrimination index, meaning success on the exam and the item seem to be uncorrelated, so the item probably isn’t discriminating well between high and low performers. This is suspicious and suspicions are reinforced by the fact that the upper & lower 27% of the performers scored almost the same on the item. To investigate, we can also look at the percent correct and see it was 99%, which is really high and may mean the item’s too easy. Looking at response frequencies, many answer choices weren’t believable for students because they didn’t pick them. Our advice would be to review the full text of the item and response options to edit these for higher plausibility. Don’t hesitate to contact the Instructional Design and Assessment team for more information on exam tagging, item statistics, and other assessment or instructional design topics. And don’t forget to visit bit.ly/msucomexams to check out our job aids on exams and other topics.