Revisiting the Effect of Item Purification on Differential Item Functioning; Real Data Findings

Author:

Year-Number: 2018-Volume 10, Issue 5
Yayımlanma Tarihi: null
Language : null
Konu :
Number of pages: 139-147
Mendeley EndNote Alıntı Yap

Abstract

Keywords

Abstract

One of the important issues facing practitioners in the process of determining Differential Item Functioning (DIF) in a test, is that the presence of one or more items with DIF may affect the results of determining DIF in other items. In this case, items that do not function differentially could be falsely identified as showing DIF, which leads to an undesirable increase in Type I error. As a solution to this problem, it has been proposed that the items showing DIF are iteratively excluded from the analyses, in a process called item purification. The purpose of this study is to compare the results of gender based DIF analyses when item purification is conducted and when it is not. The study group consisted of 655 students who take undergraduate course of Measurement and Evaluation at a state university in İstanbul. The data collection tool consisted of 25 multiple choice items covering the curriculum in the course. Data analyses were performed using the R statistical program. The "difR" package was used for this analysis. The Mantel-Haenszel, Standardization, Logistic Regression, Lord's Chi-Square, Raju and Breslow Day methods were used during the purification and non purification processes. The findings showed that the DIF results with and without the iterative processes were changed and the numbers of DIF items showed difference.

Keywords