That's a high-level explanation of the discrimination index and how you can use the raw data math to get a sense of how the question performed across good student versus bad student. But suppose this question is used in more than one exam, what you can also do is you can click on the question and click on View Normatic Data, and what you have here is you actually have a further expansion of how the question is selected and also across all the product how the discrimination index behaves. As you can see, only two exams have enough data to compute the discrimination index. The sample size represents typically 27% of all the users. So this exam, we have 511 sample size. That means that represents that we're using 27% of the submissions. So that's the top versus bottom student, whereas this exam has only 28 data points for discrimination index calculation. And now you can also see that the discrimination index for each exam is going to vary a little bit because the user answer the exam is different, and the sample size is also different. In addition, you can see that the discrimination index can be calculated based on all the user taking the question regardless which exam is taken in. So in this example, it's sort of average of the two, but you can also have a weighted average in the sense that we're going to give the discrimination index with the higher sample size more weight. So this is the weighted average. So there's just a few different ways to look at the discrimination index when the question is being reused across multiple exams, especially if those exams have a very distinct audience that is taking the exam. So the OASIS definitely calculate all this on the fly so that you can look at it and just sort of subjectively decide how to understand the data.

discrimination index

question performance

normative data

exam analysis

OASIS software

Question Rating