Browsing by Author "Hastings, C. Nicholas"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Group differentiated prediction
    (1984) Linn, Robert L.; Hastings, C. Nicholas
    Studies of predictive bias have frequently shown that a prediction equation based on majority group members tends to overpredict the criterion performance of minority group members. Two statistical artifacts that may cause the overprediction finding are reviewed and evaluated using data for black and white students at 30 law schools. It is shown that (1) the degree of overprediction decreases as the predictive accuracy for white students increases, and (2) that overprediction can be caused by the effects of selection on variables not included in the regression model. Use of Heckman’s (1979) procedure to adjust the estimates of the regression parameters was found to essentially eliminate overprediction.
  • Thumbnail Image
    Item
    Item bias in a test of reading comprehension
    (1981) Linn, Robert L.; Levine, Michael V.; Hastings, C. Nicholas; Wardrop, James L.
    The possibility that certain features of items on a reading comprehension test may lead to biased estimates of the reading achievement of particular subgroups of students was investigated. Eight nonoverlapping subgroups of students were defined by the combinations of three factors: student grade level (fifth or sixth), income level of the neighborhood in which the school was located (low and middle or above), and race of the student (black or white). Estimates of student ability and item parameters were obtained separately for each of the eight subgroups using the three-parameter logistic model. Bias indices were computed based on differences in item characteristic curves for pairs of subgroups. A criterion for labeling an item as biased was developed using the distribution of bias indices for subgroups of the same race that differed only in income level or grade level. Using this criterion, three items were consistently identified as biased in four independent comparisons of subgroups of black and white students. Comparisons of content and format characteristics of items that were identified as biased with those that were not, or between items biased in different directions, did not lead to the identification of any systematic content differences. The study did provide strong support for the viability of the estimation procedure; item characteristics, estimated with samples from different populations were very similar. Some suggestions for improvements in methodology are offered.