NotesFAQContact Us
Search Tips
ERIC Number: ED328578
Record Type: Non-Journal
Publication Date: 1991-Jan
Pages: 26
Abstractor: N/A
Reference Count: N/A
A Brief Overview of Three Classes of Methods for Detecting Item Bias.
Fisk, Yvette Hester
The reasons for recent endeavors to evaluate item bias are discussed, and item bias is defined. Some of the literature regarding the most promising methods of detecting item bias is reviewed. Three classes of methods for detecting item bias are discussed using concrete examples and illustrations. These methods are: (1) latent trait; (2) chi-square; and (3) item difficulty methods. The item difficulty method is considered to be the least computationally demanding method of the three. A small data set consisting of two groups of 17 examinees each taking a 15-item dichotomously-scored test is provided to illustrate the delta-plot technique--a user-friendly method involving item difficulties. It is concluded that researchers must decide on the best method to use based on: (1) the application of the results; (2) the level of statistical sophistication required; and (3) the practicality of implementing each method. Latent trait methods are statistically superior to chi square and item difficulty methods; however, the latter two methods are easier to implement. A 28-item list of references, two tables, and four graphs are included. (Author/SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A