NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED520494
Record Type: Non-Journal
Publication Date: 2010
Pages: 17
Abstractor: ERIC
Reference Count: 34
ISBN: N/A
ISSN: N/A
Correcting Correlations When Predicting Success in College
Saupe, Joe L.; Eimers, Mardy T.
Association for Institutional Research (NJ1), Paper presented at the Annual Forum of the Association for Institutional Research (50th, Chicago, IL, May 29-Jun 2, 2010)
The purpose of this study is to illustrate techniques for correcting a correlation between a predictor of success in college (admission test score or indicator of high school performance) with a measure of success in college (one-year retention or first-year GPA) given the restricted variances in the population used to calculate the correlations. In other words, this study demonstrates procedures for estimating correlations in the unrestricted population (students who attend college and students who do not attend college) based upon correlations calculated for the restricted population (students who attend college). A secondary purpose is to set the foundation for and stimulate additional studies designed to estimate these correlations in other unrestricted higher education and college student populations. This study focuses on correlations involving admission test scores, indicators of success in high school, and first-year college GPA. The data for this study come from a population of first-time freshmen who entered a major research university with moderately selective admission standards in the fall 2008 semester, whose high school class percentile rank was 50 or greater, who entered the fall semester as full-time, degree-seeking students, and who completed both semesters with complete data for the study variables. There are 3,668 students in this population. In sum, the true relationships between predictor variables and college success measures can be masked by restricted range as well as other extraneous variables. The present study demonstrates the influence that restricted range can have on this relationship and suggests that these predictor variables are probably more accurate than what is generally shared in the literature and in practice. This study will have been successful if it stimulates others to explore the use of the correction formulas to estimate correlations between predictor variables and indicators of success in college for unselected populations. (Contains 3 tables.)
Association for Institutional Research. 1435 East Piedmont Drive Suite 211, Tallahassee, FL 32308. Tel: 850-385-4155; Fax: 850-383-5180; e-mail: air@airweb.org; Web site: http://www.airweb.org
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: High Schools; Higher Education
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: Association for Institutional Research