NotesFAQContact Us
Collection
Advanced
Search Tips
Peer reviewed Peer reviewed
Direct linkDirect link
ERIC Number: EJ930636
Record Type: Journal
Publication Date: 2010
Pages: 11
Abstractor: ERIC
Reference Count: 23
ISBN: N/A
ISSN: ISSN-0017-8055
On Replicating Ethnic Test Bias Effects: The Santelices and Wilson Study
Freedle, Roy O.
Harvard Educational Review, v80 n3 p394-404 Fall 2010
In this commentary, the author discusses two recent replications (Santelices & Wilson, 2010; Scherbaum & Goldstein, 2008) of some of his earlier work on SAT items using the differential item functioning (DIF) statistic wherein he contrasted the test performance of African American examinees with White examinees (Freedle, 2003). In this commentary, it becomes clear how each of these two replications covers different aspects of the DIF statistic and addresses criticisms of his work. These replication studies find, in general, that whenever a significant ethnic correlation occurs between DIF and item difficulty, regardless of the item type, the African American group is found to perform differentially better on hard items than White examinees of matched ability and differentially worse than White examinees on easy items. As Scherbaum and Goldstein (2008) claim, this DIF correlation represents a counterintuitive finding, given the test bias found in previous studies (Freedle, 2003, p. 2; Gould, 1995). The author briefly advances a hypothesis explaining why this DIF correlation result is so counterintuitive. While Santelices and Wilson's (2010) findings for the verbal SAT items replicate the general pattern of DIF results the author reported (Freedle, 2003), there are some differences between these two studies when one focuses on particular item types. The author focuses on these differences because he believes they will put researchers on a path toward building a deeper theoretical understanding of the sources of test bias, at least at the individual item level. He also reviews the Scherbaum and Goldstein (2008) DIF replication study that used non-SAT content to address the assertion that his significant DIF findings (Freedle, 2003) represent a statistical artifact (see Wainer & Skorupski, 2005). Finally, he addresses some strategies for reducing mean test score differences between minority and White examinees. (Contains 2 notes.)
Harvard Education Publishing Group. 8 Story Street First Floor, Cambridge, MA 02138. Tel: 617-495-3432; Fax: 617-496-3584; e-mail: hepg@harvard.edu; Web site: http://www.gse.harvard.edu/hepg/her.html
Publication Type: Journal Articles; Opinion Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A
Identifiers - Assessments and Surveys: SAT (College Admission Test)