Date Approved

2019

Degree Type

Open Access Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Leadership and Counseling

Committee Member

Ronald Williamson, Ed.D.

Committee Member

David Anderson, Ed.D.

Committee Member

Beth Kubitskey, Ph.D.

Committee Member

Jaclynn Tracy, Ph.D.

Abstract

The number of applications to postsecondary institutions continues to increase year over year, and in most cases, the number of applications exceeds the number of students admitted. The use of standardized tests continues to grow to help in these admissions decisions. Due to both high usage rates and the changing demographics of our nation’s student population, the study of test bias is still a relevant conversation today. Aside from larger issues of equity and access, particularly in STEM courses, this has implications for leaders in higher education because universities have a stake in ensuring students are academically fit for the curriculum. The goal of this study was to understand whether differential prediction existed in the ACT science subscore and the extent to which this index was differentially valid for individuals of various demographic subgroups regarding their performance in introductory science courses. Differential prediction and differential validity were investigated through quantitative methods whereby correlation coefficients and regression models were determined for different subgroups based on various demographic characteristics. Results of this study revealed variable differential validity across gender, ethnicity, and student major. Results also revealed differential prediction completely across ethnicity and student major, and variably across gender, Pell-eligibility, and first generation status. In many cases, the results were consistent with current literature in that female student performance was often underpredicted whereas non-White student performance was often overpredicted. In terms of fairness, test bias was highly dependent on particular predictor/criterion combinations for each demographic subgroup investigated. Criterion bias was predominant and presented as a considerable concern as well.

Share

COinS