Author ORCID Identifier

Date Available


Year of Publication


Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation


Public Health


Epidemiology and Biostatistics

First Advisor

Dr. Philip M. Westgate


Interpersonal violence such as teen dating violence is a severe public health problem. Teen dating violence, including sexual violence (unwanted sexual contacts or activities), physical and psychological dating violence, sexual harassment, and stalking, affects high school students' physical and mental health and academic achievement in the United States. Dating violence is linked to psychological abuse perpetration in the future, depression, anxiety, and hostility. The teen dating violence victimization experience was related to antisocial behavior, drug abuse, increased heavy drinking, depression, suicidal ideation, smoking, and adult interpersonal violence victimization during adolescence. The detrimental effects of interpersonal violence demonstrate the critical importance of prevention. The bystander-based prevention approaches have been recognized as successful in providing strategies for violence prevention. This program teaches individuals to identify potentially violent situations and intervene safely to prevent violence.

Cluster randomized trials are often utilized to design a study implementing a bystander-based intervention in school settings. In a cluster-randomized trial, groups of individuals are randomized instead of individuals themselves. For example, as opposed to individual students, schools were randomized in the Green Dot High School study because the Green Dot intervention needs to be applied to the entire school. Enrolled schools were followed for five years. As we anticipate future studies on interpersonal violence in school-based settings, questions remain on design features such as intra-cluster correlation coefficients for different dating violence outcomes. When analyzing data from a longitudinal cluster-randomized study, such as Green Dot High School, individual students' behavioral outcomes cannot be linked over time. Therefore, the question remains on how best to analyze the behavioral outcomes measured from such studies improving the power of the analysis without violating students' anonymity. Recent studies suggest that more prolonged exposure to prevention training leads to better outcomes. However, resources are often lacking to implement these programs for such an extended period. Therefore, questions remain on how much exposure to these types of intervention programs is needed.

In the first study, to help researchers design future similar school randomized trials, we provided design features such as observed effect sizes, intra-cluster correlation values for sexual and dating violence outcomes, and school size information. In the second study, we propose an alternative way to analyze data from cluster randomized trials in high-school settings to improve power when testing for intervention-time interaction effects and main effects of grade. Furthermore, we analyzed data from the Green Dot High School setting investigating the effects of different dosage levels on perpetration and victimization of dating violence outcomes in the last study. The questions addressed in this dissertation will provide researchers a framework for future design and analyses of school-based cluster randomized trials.

Digital Object Identifier (DOI)

Funding Information

The research was funded through Teaching and Research Assistantship provided by the University of Kentucky (2017-2021).