The Study of the Effect of Anchor Items Showing or Not Showing Differantial Item Functioning to Test Equating Using Various Methods


Demirus K. B., GELBAL S.

JOURNAL OF MEASUREMENT AND EVALUATION IN EDUCATION AND PSYCHOLOGY-EPOD, cilt.7, sa.1, ss.182-201, 2016 (ESCI) identifier

Özet

The purpose of this study to compare the results of equating methods based on Item Response Theory when all the anchor items showing or not showing gender based uniform DIF. In the study the effect of DIF items on test equating presented on real data with horizontal equating using seperate calibration methods and equivalent groups with anchor test design. DIF analysis conducted on EASDIF software for "Mantel-Haenszel" method and on SPSS with syntax presented by Zumbo for "logistic regression" method. For test equating methods "mean-mean", "mean-sigma", "Haebara" and "Stocking-Lord" were used. The performances of the equating methods were evaluated through RMSD equating errors based on difference of ability estimates. BILOG-MG was utilized for the prediction of item parameters and ability, IRTEQ software was utilized for test equating. Data set for the study was obtained from the forms of science test which applicated to 1350 students in 8th grade. According to the results of the study when the anchor items with uniform DIF favored males were used for equating, mean-mean method produced the biggest equating error whereas mean-sigma method produced the smallest. When the anchor items with no-DIF were used for equating the biggest equating error was obtained from mean-sigma method and smallest equating error was obtained from Stocking-Lord ve Haebara methods in equal to each other.