An Analysis of Differential Bundle Functioning in Multidimensional Tests Using the SIBTEST Procedure

Creative Commons License


INTERNATIONAL JOURNAL OF ASSESSMENT TOOLS IN EDUCATION, vol.9, no.2, pp.319-336, 2022 (ESCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 9 Issue: 2
  • Publication Date: 2022
  • Doi Number: 10.21449/ijate.946978
  • Journal Indexes: Emerging Sources Citation Index (ESCI), ERIC (Education Resources Information Center), TR DİZİN (ULAKBİM)
  • Page Numbers: pp.319-336
  • Keywords: Differential item functioning, Differential bundle functioning, Multidimensionality, SIBTEST, Type 1 error, Power rate, ITEM BIAS, MONTE-CARLO, MODEL
  • Hacettepe University Affiliated: Yes


This study aims to analyze the differential bundle functioning in multi-dimensional tests with a specific purpose to detect this effect through differentiating the location of the item with DIF in the test, the correlation between the dimensions, the sample size, and the ratio of reference to focal group size. The first 10 items of the test that is comprised of 30 items were acknowledged as the bundle. The data in line with the parameters were generated via SAS program as two categories (1-0) and multidimensional through an extended 2PL model. Differential bundle functioning was detected via the SIBTEST procedure. The results of the study were interpreted according to the criteria of the power rate and the type I error. When the results were reviewed, the analysis of the bundle revealed that the more the correlation between the two dimensions increased, relatively the less the power rates became. It was observed that the power rates, which were obtained according to two different sample sizes in the study, increased as the sample size increased. Another result as to the SIBTEST's power for detecting DIF was the highest when the ratio of reference to focal group size was equal. According to the results of the type I error rate, the error rate was observed to be relatively decreasing as the correlation between the dimensions increased and it was observed to be increasing as the sample size increased. Also, the highest error rate was obtained when the ratio of the samples was equal.