ERIC Number: ED181029
Record Type: RIE
Publication Date: 1978
Reference Count: 0
Gardner, Eric F.
NCME Measurement in Education, v9 n3 p 1-5 Sum 1978
It is suggested that bias--when associated with a predictor, a test, or a statistical estimator--is not always bad, in spite of the immediate negative response evoked by the word, bias. Four settings are described to illustrate situations in which a procedure should not be summarily rejected due to bias: (1) educational researchers rejected the use of critical ratios for hypothesis testing, in favor of the t-test, to avoid bias which was negligible for most sample sizes; (2) ridge regression, which uses biased estimators, provides better prediction with smaller errors than do the less biased least squares estimators; (3) a utility model is proposed for competitive selection problems because it requires an explicit public statement of utilities (biases) for each subpopulation; and (4) low scores on an achievement test by a particular group do not necessarily prove that the test is biased against that group, when the test has appropriate content validity. It is concluded that bias should be recognized and explicitly considered when choosing the best solution from different alternatives, rather than pretending bias doesn't exist, fearing bias irrationally, or introducing other unknown biases in opposition. (GDC)
Publication Type: Reports - Research; Journal Articles; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: National Council on Measurement in Education, East Lansing, MI.
Note: Paper presented at the Annual Meeting of the National Council on Measurement in Education (Toronto, Ontario, Canada, March, 1978)