Electronic Theses and Dissertations

Date

2025

Document Type

Thesis

Degree Name

Master of Science

Department

Public Health

Committee Chair

Meredith Ray

Committee Member

Hongmei Zhang

Committee Member

Patricia Cisarik

Abstract

This study evaluates common approaches to analyzing ocular longitudinal data. These approaches range from univariate multiple linear regression to mixed-effect modeling to bivariate outcome models, and vary in both overall complexity, and the amount of data used for analysis. Through comparisons of relative model precision and accuracy, the study highlights how modeling choices may affect inference in the context of ocular data analysis. Through the use of simulated datasets based on reference data from a previous longitudinal study of myopia, a total of ten modeling approaches are assessed. The data are simulated under four different correlation and effect size scenarios to examine the robustness of modeling approaches and to provide more generalizable results. These simulations consistently show that more complex mixed modeling approaches result in increased model precision, relative to models with only fixed effects. The benefits of three-level mixed models compared to two-level mixed models depend on if the predictor of interest is measured at the level of the subject or the eye. As correlation between outcomes is decreased, either between the left and right eye, or within an eye over time, mixed modeling approaches show further gains in precision. Notably, bivariate outcome modeling does not outperform univariate outcome mixed models in most scenarios. However, if effect sizes are not consistent between the left and right eye, bivariate outcome modeling is able to more accurately detect and estimate these different effect sizes compared to all other evaluated modeling approaches. These results may allow for more efficient analysis of ocular data, and may enable investigators to use fewer resources to reach the same conclusions.

Comments

Data is provided by the student.

Library Comment

PDF

Notes

Open access.

Download

Share

COinS