Title: DIF items effects on Kuder-Richardson 20 formula and construct validity: an exploration using the Mantel-Haenszel method

Authors: Mahmoud Alquraan

Addresses: Department of Counseling and Educational Psychology, College of Education, Yarmouk University, Irbid, Jordan

Abstract: This study aims to explore the effect of DIF items detected by Mantel-Haenszel (MH) method on the reliability (KR20) and construct validity (internal structure) of test scores. The current study used a verbal ability test developed by Enab (1994) for Jordanian students of the age group (9-12) years. The test consists of 40 multiple-choice items distributed to seven subtests, and it was distributed to (758) students. Results show that removing DIF test items does not have a statistical effect on the reliability coefficient (KR20). The confirmatory factor analysis provides evidence that removing DIF items detected by MH leads to better construct validity (internal structure) of the test scores.

Keywords: Mantel-Haenszel method; differential item functioning; DIF; construct validity; reliability; test scores; verbal ability tests; Jordan; confirmatory factor analysis; CFA.

DOI: 10.1504/IJQRE.2016.10003212

International Journal of Quantitative Research in Education, 2016 Vol.3 No.4, pp.266 - 276

Received: 11 Dec 2015
Accepted: 02 Aug 2016

Published online: 22 Feb 2017 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article