Dissertations and Theses (Open Access)
Date of Award
Spring 4-11-2024
Degree Name
Doctor of Philosophy (PhD)
Advisor(s)
WENYAW CHAN, PhD
Second Advisor
CICI X. BAUR, PhD
Third Advisor
STACIA M. DESANTIS, PhD
Abstract
The continuous-time Markov chain (CTMC) model and latent clustering models are commonly used to study longitudinal measures of categorical outcomes. Because of its simple but powerful Markovian property, CTMC models have been widely used in medical and public health researches. Due to limitations in the standard CTMC model, there have been some studies on non-homogeneous continuous-time Markov chain (NH-CTMC) that utilized time-dependent rates, but the progresses have been limited. NH-CTMC can be more powerful than CTMC by its default nature of time-dependent rate that can be fitted to a wider range of applications in medical studies. In this study, we have derived different types of NH-CTMC models and class clustering of latent chains: two-state full ergodic model and three-state full ergodic model. We demonstrated the significance of the models and how they can be adopted in public health research and applications, specifically in longitudinal discrete outcome studies where the rates of state transitions are time-dependent and vary by subject groups.
Recommended Citation
Chang, Joonha, "LATENT CLASSIFICATION OF MULTI-MODEL NON-HOMOGENEOUS CONTINUOUS-TIME MARKOV CHAINS" (2024). Dissertations and Theses (Open Access). 267.
https://digitalcommons.library.tmc.edu/uthsph_dissertsopen/267
Included in
Biostatistics Commons, Categorical Data Analysis Commons, Community Psychology Commons, Health Psychology Commons, Longitudinal Data Analysis and Time Series Commons, Probability Commons, Public Health Commons, Statistical Theory Commons