The Effects of Educational Technology Usage Profiles and Legally Protected Bio-Demographic Data on Behaviorally-Based Predictive Student Success Models in Learning Analytics: An Exploratory Study
Abstract
In the 21st century, attainment of a college degree is more important than ever to achieve economic self-sufficiency, employment, and an adequate standard of living. Projections suggest that by 2020, 65% of jobs available in the U.S. will require postsecondary education. This reality creates an unprecedented demand for higher education, and postsecondary education enrollment has increased 48% in the last 20 years. At the same time, the U.S. concurrently is facing a crisis of underprepared college students. Nationally, retention rates in higher education have taken a downward trajectory and time to graduation has increased. At least 40% of degree-seeking students in this country never make it to graduation. This study built on the work of Campbell (2007) and Fritz (2016) by leveraging data available in learning management systems to create predictive student success models. Grounded in the classic retention theory of Astin, Tinto, and Pascarella, this study integrated elements of connectivism to form a framework for learning analytics application. The intent of this study was to examine the viability of using sociotechnical elements as a means of improving the predictive power of a set of previously established independent variables on predicting student success, while lessening the impactful of static, legally protected bio-demographic variables. This study utilized binary logistic regression and Kovanović et al.’s (2015) educational technology usage profiles to explore course level application of student success models. Course level models without legally protected variables performed equally as well when compared to institutionally derived models which included ethnicity and sex as predictors. Additionally, course level models demonstrated more actionable outcomes than institutional models did. The results of this research informed a series of operational and strategic implications for institutional consideration when exploring learning analytics as a solution to retention and student success issues.
Degree
Ph.D.
Advisors
Newby, Purdue University.
Subject Area
Educational technology|Higher education
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.