Title

An examination of selection bias in the Florida College Level Academic Skills Test

Date of Award

1988

Availability

Article

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Educational Research

First Committee Member

Maria M. Llabre, Committee Chair

Abstract

The purpose of this study was to examine the extent to which selection bias is present in the use of the Florida College Level Academic Skills Test (CLAST), given three different perspectives and three sets of passing standards. Although the CLAST is not designed to be a predictor of future performance, students in the state of Florida are required to pass all four CLAST subtests in order to receive an Associate in Arts degree or to be admitted into upperclass standing. Analyses were based on separate regressions of first term state university system GPA on CLAST subtest scores. The sample consisted of 6330 students who enrolled in or graduated from Miami-Dade Community College between 1980 and 1984. White Non-Hispanics accounted for 32.58% of the students, while 7.95% were Black Non-Hispanic, 57.69% were Hispanic, and 1.79% were Foreign Nationals.Cleary's regression model revealed selection bias (unequal intercepts) in all analyses but one. Use of the overall regression equations overpredicted Black Non-Hispanic GPA by 0.20-0.47 points, underpredicted White Non-Hispanic GPA by 0.03-0.07 points, and underpredicted Hispanic GPA by 0.01-0.07 points. Prediction was improved by the inclusion of lowerclass GPA in the regression equation. Thorndike's constant ratio model showed that the proportion of students successful on the exam to the proportion with successful upperclass performance was not equal across ethnic groups. Bias was indicated in favor of Black Non-Hispanics at the 1984 and 1986 standards, and against this group at the 1989 standards. Bias was indicated against Hispanics on the reading and essay subtests given 1984 and 1986 standards, and on all four subtests given 1989 standards. Einhorn and Bass's equal risk model showed significantly fewer (10 to 14 percent) White Non-Hispanics than Black Non-Hispanics predicted to have unsuccessful upperclass performance given performance at all cutscores on all subtests. Similarly, significantly fewer (2-4 percent) White Non-Hispanics than Hispanics were predicted to have unsuccessful upperclass performance given performance at the cutscore on the computation, writing, and essay subtests at 1984 and 1986 standards, and on the writing and essay subtests at the 1989 standards.

Keywords

Education, Tests and Measurements

Link to Full Text

http://access.library.miami.edu/login?url=http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqdiss&rft_dat=xri:pqdiss:8827961