Abstract
Detection of differential item functioning (DIF) is needed in the development of tests to obtain useful items. The Mantel-Haenszel method and standardization are tools for DIF detection based on classical theory assumptions. The study was conducted to highlight the sensitivity and accuracy between the Mantel-Haenszel method and the standardization method in DIF detection. Simulation design (a) test participants consisted of 1000 responses in both the reference and focus groups, (b) the size of the proportion of DIF (0.1; 0.25; 0.50; and 0.75), and (c) the length of the multiple choice test with 40 choices the answer. Research shows that the Mantel-Haenszel method has the same sensitivity as the standardization method in DIF proportions of 10% and 25%, however, when the ratio of DIF proportions above 25% the standardization method is less sensitive, and conversely the sensitivity of the Mantel-Haenszel method increases. The standardization method has higher accuracy than the Mantel-Haenszel method in the DIF proportion of 10%, however, when the size of the DIF proportion above 10% the accuracy of the standardization method decreases, the accuracy of the Mantel-Haenszel method is higher than the standardization method. Thus, if the ratio of DIF is detected by the standardization method of (≤0.10), then the results of the standardization method are preferred as a reference. Conversely, if the proportion of DIF detected by the standardization method is (≥0.10), then the result of the Mantel-Haenszel method is chosen as a reference.
Highlights
The measurement process through tests aims to obtain information about the extent to which students’ achievements or abilities are after the learning process
The Mantel-Haenszel method has the same sensitivity as the Standardization method in differential item functioning (DIF) proportions of 10% and 25%, but when the size of the DIF proportion increases causing the standardization method to be less sensitive, the Mantel-Haenszel method is more sensitive when the DIF proportion is above 25%
The Standardization method has higher accuracy than the Mantel-Haenszel method in the DIF proportion of 10%, but when the size of the DIF proportion increases which is above 10% the accuracy of the standardization method decreases, so the accuracy of the Mantel-Haenszel method is higher than the Standardization method
Summary
The measurement process through tests aims to obtain information about the extent to which students’ achievements or abilities are after the learning process. The test package used consists of several items as a test tool to measure students’ abilities. The tests developed can be formative and summative tests, class-based tests and others. Presented both in the form of dichotomy and polytomy. Tests in the form of dichotomies are like multiple choice questions, where the test only consists of two models of answers which are correct with a score of one “1” and wrong with a score of “0” Whereas the test is in the form of a polytomy, which is like a description question, where students’ answers are assessed based on the assessment criteria with the specified score range
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
