您的浏览器禁用了JavaScript(一种计算机语言,用以实现您与网页的交互),请解除该禁用,或者联系我们。[ACT]:Predictiing College Grades From ACT Assessment Scores and High School Course Work and Grade Information - 发现报告
当前位置:首页/行业研究/报告详情/

Predictiing College Grades From ACT Assessment Scores and High School Course Work and Grade Information

文化传媒2014-09-12ACT野***
Predictiing College Grades From ACT Assessment Scores and High School Course Work and Grade Information

A C T Research Report Series91-3Predicting College Grades From ACT Assessment Scores and High School Course Work and Grade InformationJulie P. IMoble For additional copies write: ACT Research Report Series P.O. Box 168 Iowa City, Iowa 52243©1991 by The American College Testing Program. All rights reserved. PREDICTING COLLEGE GRADES FROM ACT ASSESSMENT SCORESAND HIGH SCHOOL COURSE WORK AND GRADE INFORMATIONJulie P. Noble ABSTRACTThis study examined the accuracy of predictions of English, mathematics, social studies, and natural sciences course grades, and of overall freshman GPA. The predictions were based on ACT Assessment test scores and on high school course work and grade information from the ACT Assessment Course Grade Information Section (CGIS). Estimates of prediction accuracy were compared to those obtained using ACT scores and the four self-reported grades from the registration folder (TH index), ACT Assessment scores alone, and CGIS information alone as predictors.Base-year prediction models were developed using student records from the 1986-87 and 1987-88 Prediction Research Services history files; these models were then crossvalidated using data from the same institutions for 1988-89. Separate models were developed and crossvalidated for juniors and seniors. In addition, total group prediction models were developed and crossvalidated separately for juniors and seniors.The results showed that most ACT/CGIS models slightly increased prediction accuracy in some subject areas over that obtained by the TH index. The model based on the four ACT scores and an average of 23 grades improved prediction accuracy over that of the TH index for more than 50% of the institutions. The amount of improvement was modest, however. The results clearly supported the use of prediction models based jointly on ACT scores and high school grades, rather than on either ACT scores or grades alone. Julie P. NobleCollege admissions or placement decisions are often based in part on predictions of students' performance during their freshman year (e.g., course grades or GPA). Students whose predicted performance falls above a certain level of performance are admitted into the college or course; students whose predicted performance falls below the specified level may be denied admission or may be admitted under special conditions.The Prediction Research Services (formerly Standard Research and Basic Research Services) provided by ACT allows institutions to develop predictions of students' grades in specific college courses. The predictions are based on regression models composed of students' ACT test scores (in English, mathematics, social studies, and natural sciences; in English and mathematicsduring the transition to the enhanced ACT Assessment) and their self-reportedhigh school grades in the same subject areas.Predicted grades and regression weights in the Prediction Research Services are derived for each institution using the TH index, which is calculated using two prediction models:(1) Yt = aQ + a1 * ACT English Usage score+ a2 * ACT Mathematics Usage score+ a3 * ACT Social Studies Reading score+ a4 * ACT Natural Sciences Reading score(2) YH = b0 + b., * HS English grade+ b2 * HS Mathematics grade+ b3 * HS Social Studies grade+ b4 * HS Natural Sciences gradeFor these equations, YT and YH are the predicted course grades for the two models; the subscript T refers to the test models, or T index, and the H refers to the high school grade model, or H index. The TH index is the average of the twc college grade predictions, the T index and the H index.The TH index resembles an 8-variable prediction model and has been shown to yield predictions of comparable accuracy (ACT, 1965). The values a0, a.,, a2, a3, a4, b0, b1f b2, b3, and b4 are regression weights; all values are specific to an institution and the course grade being predicted. The weights arePREDICTING COLLEGE GRADES FROM ACT ASSESSMENT SCORESAND HIGH SCHOOL COURSE WORK AND GRADE INFORMATION calculated from the college course grades, ACT test scores, and self-reported high school grades for students from each institution participating in the Prediction Research Services.The self-reported grades in the four major subject areas are those traditionally collected on the ACT Assessment Registration Folder (RF). There are several limitations in using the RF grades to predict college grades: first, a postsecondary institution cannot determine the exact content of the courses taken in a particular subject area, and thus cannot determine their appropriateness for predicting college course grades. In addition, the institution is limited to high school grades for courses in the four major subject areas (English, mathematics, social studies, natural sciences). The Registration Folder does not include course work in foreign languages or fine arts. Further,