This dissertation research focuses on assessing student behavior, academic emotions, and knowledge within a middle school online learning environment, and analyzing potential effects on students’ interests and choices related to decisions about going to college. Using students’ longitudinal data ranging from their middle school, to high school, to postsecondary years, this dissertation uses quantitative methodologies to investigate antecedents to college attendance that occur as early as middle school. The dissertation asks whether student behavior, academic emotions, and learning as early as middle school can be predictive of college attendance years later. This is investigated by developing predictive and structural models of said outcomes, using assessments of learning, emotions and engagement from student interaction data from an online learning environment they used in their middle school curriculum. The same middle school factors are also assessed with self-report measures of course choices, interests in college majors and careers formed when they were in high school. The dissertation then evaluates how student choices and interests in high school can mediate between the educational experiences students have during middle school and their eventual college attendance, to give a fuller illustration of the cognitive and non-cognitive mechanisms that students may experience throughout varied periods in school. Such understanding may provide educators with actionable information about a students’ in-depth experiences and trajectories within the college pipeline.

Short-to-medium projections were made of student semester credit hour enrollments for North Texas State University and the Texas Public and Senior Colleges and Universities (as defined by the Coordinating Board, Texas College and University System). Undergraduate, Graduate, Doctorate, Total, Education, Liberal Arts, and Business enrollments were projected. Fall + Spring, Fall, Summer I + Summer II, Summer I were time periods for which projections were made. A new regression analysis called "cycle regression" which employs nonlinear regression techniques to extract multifrequential phenomena from time-series data was employed for the analysis of the enrollment data. The heuristic steps employed in cycle regression analysis are similar to those used in fitting polynomial models. A trend line and one or more sin waves (cycles) are simultaneously estimated using a partial F test. The process of adding cycle(s) to the model continues until no more significant terms can be estimated.