ERIC Number: ED157942
Record Type: RIE
Publication Date: 1978-Mar
Reference Count: 0
An Evaluation of Select Approaches For Biased Item Identification.
Rudner, Lawrence M.; Convey, John J.
Transformed item difficulties, chi-square, item characteristic curve (icc) theory and factor score techniques were evaluated as approaches for the identification of biased test items. The study was implemented to determine whether the approaches would provide identical classifications of items as to degree of aberrance for culturally different populations and classifications of minimal bias for subsamples of a single population. Actual item response data were obtained from 2,637 hearing impaired and 1,607 normal students, and two pseudo-culture group samples of subjects from the same population with different mean total scores. The Stanford Achievement Test, a 48-item test of reading comprehension, was administered to the students. In the diverse culture group comparison, subjects responded to a pool of items which measured reading comprehension, and several items were found to be biased. Degrees of aberrance in the equal-culture group were consistently low for the icc theory and chi-square approaches. These approaches were felt to be the most promising, and the icc theory approach was sensitive to individual and group bias. The factor score and chi-square approaches were inadequate for identifying bias. Several common items were identified for the diverse culture comparison by the transformed item difficulties, icc theory, and chi-square approaches. Recommendations are made for future studies incorporating known parameters and distractor response analysis. (Author/JAC)
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: N/A
Identifiers: Chi Square
Note: Paper presented at the Annual Meeting of the American Educational Research Association (62nd, Toronto, Ontario, Canada, March 27-31, 1978)