Predicting Student Success at a Large State University using Multiple Linear Regression and Hierarchical Clustering
Date
2022-08
Authors
Collins, Abigail
Major Professor
Genschel, Ulrike
Advisor
Committee Member
Hofmann, Heike
Ommen, Danica
Journal Title
Journal ISSN
Volume Title
Publisher
Altmetrics
Abstract
As a result of the on-going Covid-19 pandemic post-secondary institutions across the US have begun to drop requiring standardized tests such as the ACT or SAT exam for the purpose of admitting students. However, many institutions have previously relied on these assessments to predict student success. In some instances, ACT or SAT scores have been part of a carefully calibrated index to assess how likely a student is to graduate from college. Students which surpass a certain threshold are then automatically admitted to the institution. For our analysis, we consider data from a large state university having relied on such an index, and we investigate alternative ways to predict retention rates and, thus,
student success. We use multiple linear regression to predict the ACT score itself using an 11th grade state-wide assessment score that is available for all students who attended high school in the state the university is located in. The predicted
ACT scores were subsequently used in place of the actual ACT to accurately predict the University’s admissions index. Additionally, hierarchical agglomerate clustering was used to predict student retention rates after the first year. The accuracy of our predictions dependents on the data available to us, specifically if missing data can be considered missing at random or not.
Series Number
Journal Issue
Is Version Of
Versions
Series
Academic or Administrative Unit
Type
creative component
Comments
Rights Statement
Copyright
2022