The Contrasting Groups Method is a common approach to setting a cutscore. It is very easy to do, but has the important drawback that some sort of “gold standard” is needed to assign examinees into categories such as Pass and Fail. This “gold standard” should be unrelated to the test itself.
For example, suppose you wanted to set a cutscore on a practice test that is helping examinees determine if they are ready for a high-stakes certification test. You might have past data for examinees who took both your practice exam and the actual certification test. Their results from the certification test can be used to assign them to groups of Pass or Fail, and then you can evaluate the practice test score distributions for each group. These distributions are typically smoothed, and their intersection represents an appropriate cutscore for the practice test. In the example below, the two curves intersect near a score of 85, suggesting that this is an appropriate cutscore for the practice test that will closely predict the results of the official certification test.
I developed a simple tool in MS Excel that allows our psychometricians to easily produce both the smoothed and unsmoothed versions of this method, given nothing more than a list of practice test scores and “real” test classification for examinees. Click here to download the tool.
ASC hosted a webinar on June 14, 2022, as part of its AI In Assessment Series. This edition is an interview with the father of
Incremental validity is an aspect of validity that refers to what an additional assessment or predictive variable can add to the information provided by existing
