The concept of doing the erasure analysis is the same when an item can have more than one correct answer as when the item has only one correct answer. The general idea is whether the answer change resulted in a change to the student's score. If the student changed from one correct answer to another correct answer, we would say it was right-to-right answer change. The net effect on the score is zero. So, we don't care about the answer change. Similarly, we don't care about wrong-to-wrong answer changes. The general procedure has been to count the number of wrong-to-right answer changes and analyze those. If your items are scored dichotomously (I.e., correct or incorrect) there will be no difference in your analysis. If you have polytomously scored items, you will need to adapt. I haven't seen any research for ploytomously scored items and erasure analysis.
------------------------------
Dennis Maynes
Chief Scientist
Caveon Test Security
Original Message:
Sent: 02-18-2016 09:25
From: Alassane Savadogo
Subject: Psychometrics/Answer Changes on Test Item Analysis/Data Forensics
Hello everyone,
I am reaching out to the ASA community for information regarding erasure analysis. I have worked pretty much on erasure analysis using multiple choice items with only one correct answer. I am now into working on erasure using multiple choice items with more than one correct answers. I am wondering if anyone can refer me to documentation on how to handle erasure analyses on these type of items.
Best regards,
Alassane