Graduation Year

2025

Document Type

Dissertation

Degree

Ed.D.

Degree Name

Doctor of Education (Ed.D.)

Degree Granting Department

Educational Measurement and Research

Major Professor

Eunsook Kim, Ph.D.

Committee Member

John M. Ferron, Ph.D.

Committee Member

Robert F. Dedrick, Ph.D.

Committee Member

Tony X. Tan, Ed.D.

Keywords

Student engagement, Graduation, Latent variable modeling, Survey data

Abstract

College graduation is a key goal for both students and institutions, yet increasing rates of four-year graduation has proven challenging. An initiative that has been implemented focuses on enhancing student participation both in the classroom and across the broader academic setting. The National Survey of Student Engagement (NSSE) has been used by many institutions worldwide as a tool to evaluate student engagement through self-reported data. Research indicates that NSSE is linked to students’ academic success, such as better course grades, retention, and timely graduation. However, the existing relationships have been opaque and even contradictory. This study employs a newly developed analytical strategy that combines latent class analysis and propensity score weighting methods to explore the causal connection between student engagement profiles and the likelihood of graduation in four years. In addition, another primary goal of this dissertation is to demonstrate step-by-step the methodological approach and the corresponding procedures when concluding the causal connection between student engagement profiles and degree completion. It encourages researchers to replicate the current study with similar data sources to enrich the discussion on student engagement and academic success. Finally, this dissertation also compared different approaches to conducting latent class analysis and propensity score estimation and provided advice on choosing the most appropriate approaches given different research situations.

The results suggest that students in the Reflective & Integrative (RI) – Higher Order (HO) Learning strategy users’ class were 1.14 times more likely to graduate in four years than the Low strategy users. In addition, students in the RI strategy users’ class were 1.23 times more likely to graduate in four years than the HO strategy users. Lastly, students in the HO strategy users’ class were 1.28 times more likely to graduate in four years than the Low strategy users. On the other hand, the findings from the methodological comparisons pointed out that a standard latent class analysis (LCA) yielded four qualitatively different classes, whereas the latent class tree (LCT) analysis yielded 11 classes. Due to the advantage of LCT’s tree structure, all 11 classes are hierarchically linked, allowing substantive interpretation to happen at any level. Moreover, comparisons between two propensity score estimation methods of multinominal logistic regression (MLR) and neural networks (NN) analysis indicated that NN produced better covariate balance. This study also provided discussions on the substantive and methodological findings. It also gave suggestions for future research, such as incorporating fallible covariates, causal inference with latent outcome, and NN analysis with smaller datasets.

The significance of this dissertation includes informing educators, students, parents, and higher education administrators that student engagement and timely graduation are causally related. Because this dissertation examined the differential effects using latent class analysis, tailored intervention plans can be implemented for the targeted student population rather than the entire population. Secondly, the findings of this dissertation contribute to the ongoing discussions regarding the validity of NSSE and its effectiveness in assisting stakeholders in enhancing student success. The established associations between NSSE and students’ academic success are mixed and correlational in nature due to the lack of control over all available confounding variables that might provide an alternative explanation of the relationship. Thus, through the application of the propensity score weighting method for covariate adjustment, this study provides fresh evidence of the connection between student engagement and academic achievement. Lastly, this dissertation encourages researchers in higher education to replicate the analytical approaches with various data sources. This will greatly enrich the discussions on student engagement and timely graduation and scrutinize the modern methodological development of combining latent variable modeling and causal inference.

Download

Share

COinS