Date of Award
2025
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Statistics
First Advisor
Abigail K. Crocker
Abstract
This article examines how different statistical methods and data decisions influence the usability and interpretation of underrepresented populations. As a motivating example, we compare Cumulative Risk (CR), Multiple Individual Risk (MIR), and Latent Class Analysis (LCA) methods to measure the associations between Adverse Childhood Experiences (ACEs) and mental health in incarcerated individuals. This study uses cross-sectional data from the 2024 Vermont Prison Research and Innovation Network (PRIN) community-engaged surveys. With a sample size of 212 adult male incarcerated individuals, we perform an LCA analysis followed by four partial proportional odds (PPO) models using the PROMIS® Global Mental Health 2 questionnaire (GMH-2) as our outcome variable. Each PPO model contains one of the four ACE variables: continuous CR, categorical CR, MIR, and the LCA created variable.
Through LCA we opted for a four-class model with the resulting class distinctions: Low Adversity (14.15%), Moderate Household Dysfunction and Abuse/Neglect (53.77%), Moderate Household Dysfunction and High Abuse and Neglect (6.13%), and High Adversity (25.94%). When measuring the association between GMH-2 and ACEs, the MIR model produced the lowest BIC and AIC scores. For this analysis, the LCA model displayed the worst model fit. Overall LCA allows for more interpretation amongst subgroups of individuals and a measure of the ACE climate in the facility. However, with small sample sizes, MIR models are recommended if researchers wish to analyze ACEs beyond prevalence.
Language
en
Number of Pages
61 p.
Recommended Citation
Moody, Abigail Kay, "Associations Between Mental Health and Adverse Childhood Experiences Among U.S. Adult Incarcerated Individuals: A Comparative Analysis of Methodological Approaches and Data Decisions" (2025). Graduate College Dissertations and Theses. 2034.
https://scholarworks.uvm.edu/graddis/2034
