ERIC Number: ED163072
Record Type: RIE
Publication Date: 1978-Mar
Reference Count: 0
A Comparative Analysis of Several Methods of Assessing Item Bias.
Ironson, Gail H.
Four statistical methods for identifying biased test items were used with data from two ethnic groups (1,691 black and 1,794 white high school seniors). The data were responses to 150 items in five subtests including two traditional tests (reading and mathematics) and three nontraditional tests (picture number test of associative memory, letter groups-inductive reasoning, and mosaic comparisons-perceptual speed). The four statistical methods compared indices for the two groups with respect to: a normal transformation of percent passing, the point biserial item test correlation, chi square, and the area between item characteristic curves (ICC). The results for three of the methods were moderately positively correlated, but there was very little agreement between the item test correlation method and the other three. A random sample of 100 black students was matched with a sample of 100 white students on the basis of total test score, and item data were compared for percent passing in the two groups. Finally, a comparison was made of the indications of bias in traditional tests and the indications of bias in the nontraditional tests. (CTM)
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: N/A
Identifiers: Chi Square; Item Discrimination (Tests)
Note: Paper presented at the Annual Meeting of the American Educational Research Association (62nd, Toronto, Ontario, Canada, March 27-31, 1978)