Biased test items and differential validity

Research output: Contribution to journalArticlepeer-review

Abstract

Tested some effects of one type of biased test items through the use of a simple psychometric model. Particular emphasis was placed on the magnitude of the difference in majority and minority group validity coefficients, and it was hypothesized that ability distributions for minority and majority (White) groups are identical, but tests are biased. An example was tested to illustrate the consequences of a profoundly biased test, but only a small difference in validity coefficients was found. The probability of detecting this difference in a standard differential validity study or by the aggregation method of J. E. Hunter et al (see record 1979-22609-001) is near the alpha level of the significance test. Thus, it cannot be assumed that minority and majority group validity coefficients will differ when a test is biased. The effects of subtest reliability on bias are discussed. Even though the test can predict criterion performance as well for minorities as for majorities, it may be inappropriate to compare the 2 types of scores. (13 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved).

Original languageEnglish (US)
Pages (from-to)526-531
Number of pages6
JournalPsychological bulletin
Volume92
Issue number2
DOIs
StatePublished - Sep 1982

Keywords

  • biased test items & difference in majority & minority group validity coefficients

ASJC Scopus subject areas

  • General Psychology

Fingerprint

Dive into the research topics of 'Biased test items and differential validity'. Together they form a unique fingerprint.

Cite this