Detecting Item Bias Using an Item Response Theory(IRT) Model

다국어 초록 (Multilingual Abstract)

This study examined item bias occurring in a. language test which administered to two groups. The two groups could be divided into Koreans and Americans based upon their first Language background. The language test used in this paper was Word Analysis of ITBS. This intended to examine item bias caused by the background of the first language.
Item bias in the language test was explored by using an item response theory model. Specifically, Rasch model which was one-parameter logistic model among item response theory models was applied to calibrate items. The difficulty measures of items which were calibrated by the Rasch measurement helped to obtain the accurate comparison of groups by the invariance of item difficulty measures.
In calibrating items, the computer program MSCALE was used. The two groups were compared for item difficulty measures which were obtained by MSCALE. The comparison was made by the computer program IPARM using chi-square statistics. The results identified five items as having significantly large group differences. That is, the five items had different difficulty measures for the two groups.
The distractor analysis and the content analysis for the five items revealed that the linguistic background of the two groups affected item difficulties. This suggests that items can be biased against a group due to the linguistic backgrounds of groups.
These results showed that item bias caused by ethnic characteristics could be well identified by using an item response theory model. By the objective measures provided by an item response theory model, item bias can be more accurately identified, This means that an item response theory model can be a useful technique for examining item bias.
Many techniques for investigating item bias are developed, but it was a problem that they use raw scores. When raw scores are used, the findings about item bias can depend on samples or items used. Therefore, the application of item response theory is required for the objective finding.
In addition, the results demonstrated how the factors causing item bias could be explained. The explanation of the factors is more meaningful in studying item bias since the explanation of the factors can help to understand the group characteristics. The understanding of group characteristics can improve teaching and learning for the group. The examination of item bias is required not only for making tests fair but also for improving teaching and learning.

목차 (Table of Contents)

• CONTENTS
• Ⅰ. Introduction = 179
• Ⅱ. Methods = 180
• Ⅲ. Results = 182
• Ⅳ. Conclusions = 185
• REFERENCES = 186
• APPENDIX = 187