Detecting Differential Item Functioning in Multidimensional Tests: Comparing Single Items to Item Bundles Using SIBTEST

Abstract

This study aimed to compare the detection of differential item functioning between single
items and item bundles in multidimensional tests. The study applied SIBTEST in small, medium, and large samples, using datasets generated from 2,000 Nakornsritammarat sixth-grade students sitting for the National Achievement Test in the 2003 academic year. The secondary data was derived from the scores of the Grade 6 Thai Language achievement test which consists of 40 items, separated into two bundles. The first bundle contained 15 items and was structured to test knowledge, whereas the second bundle contained 25 items and was structured to assess process.

The results were as follows:
1. Among the three sample sizes, differences were found in DIF between single items
and item bundles in the multidimensional tests.

2. The differential item functioning of the entire test showed that there were 4 DIF
items (10%) in the small sample size, 13 DIF items (32.5%) in the medium sample size, and 15 DIF items (37.5%) in the large sample size.

3. The differential item functioning of individual bundles revealed that there were

4 DIF items (10%) in the small sample size, 8 DIF items (20%) in the medium sample size, and 16 DIF items (40%) in the large sample size.

4. The differential bundle functioning of the second bundle in the medium sample size
was statistically significant at the .05 level.