When the precise basis of selection effect on correlation and regression equations is unknown but can be modeled by selection on a variable that is highly but not perfectly related to observed scores, the selection effects can lead to the commonly observed "overprediction" results in studies of predictive bias. (Author/PN)

Examined how coaching affects the predictive validity and fairness of scholastic aptitude tests. A coached (n=271) and uncoached (n=95) group were compared. Comparison revealed that although coaching enhanced scores on the Israeli Psychometric Entrance Test by about 25% of a standard deviation, it did not create a prediction bias or affect…

Nine methods of smoothing double-entry expectancy tables (tables that relate two predictor variables to probability of attaining success on a criterion) were compared using data for entering students at 85 colleges and universities. The smoothed tables were more accurate than those based on observed relative frequencies. (Author/CTM)

The gains in accuracy resulting from applying any of the smoothing methods appear sufficient to justify the suggestion that all expectancy tables used by colleges for admission, guidance, or planning purposes should be smoothed. These methods on the average, reduce the criterion measure (an index of inaccuracy) by 30 percent. (Author/MV)

Differential prediction of two-year grade point average is reported for gender groups within university programs of study. Johnson-Neyman analyses indicated regions on the predictor score scales where differential prediction has practical impact for a significant portion of students. Careful selection of a subset of variables can reduce…

Using predictive validity studies of the Law School Admissions Test (LSAT) and the undergraduate grade-point average (UGPA), this study examined the large variation in the magnitude of the validity coefficients across schools. LSAT standard deviation and correlation between LSAT and UGPA accounted for 58.5 percent of the variability. (Author/EGS)

College freshmen's grade point averages at 260 colleges were predicted on the basis of multiple regression equations using the four previous classes separately to compute the equations. Predictor variables were four American College Test (ACT) scores and high school grades. Predictions remained accurate over the four-year period. (Author/CTM)

A new measure of academic performance was developed through a new application of item response theory (IRT). This new criterion, an IRT-based grade point average (GPA), was used to determine the predictive validity of certain preadmissions measures for 1,564 students admitted to Stanford University in 1982. (SLD)

Item response theory (IRT) is used to develop a form of adjusted cumulative grade point average (GPA) for use in predicting college academic performance appropriately for males and females. For 1,564 students at Stanford University (California), the IRT-based GPA was more predictable from preadmission measures than the cumulative GPA. (SLD)

Whether college grades are gender-related to traditional predictors of aptitude test scores and high-school grades was studied. Fall and spring (1985-86) enrollments in 88 introductory university courses (IUCs) were analyzed. In most of the large IUCs, no gender bias held up on cross-validation in a subsequent semester. (TJH)

This study of test bias used Black, Oriental, Chicano, and White students to answer two questions: (1) Is grade point average prediction for Chicanos and Orientals similar to prediction for Blacks and Whites? and (2) Does major field mediate ethnic differences in test performance? (Author/BW)

The reliability of a method of adjusting grade point averages for differences in college-departmental grading standards was examined, using 409 female and 518 male Dartmouth graduates. Such adjustments' effects on predictive validity of high-school grades, Scholastic Aptitude Test scores, and achievement test scores were assessed, including gender…