Joint Latent Class Modeling in Longitudinal and Survival Processes

Author:
Liu, Yue, Department of Statistics, University of Virginia
Advisors:
Liu, Lei, Public Health Sciences Admin, University of Virginia
Zhou, Jianhui, Department of Statistics, University of Virginia
Ma, Jennie, Public Health Sciences Admin, University of Virginia
Hu, Feifang, Department of Statistics, University of Virginia
Abstract:

In many clinical studies, repeated measures and survival outcomes are often both of interest. Furthermore, there could exist heterogeneous latent groups among the subjects. Forsuchheterogeneousdatawithbothlongitudinalandsurvivaloutcomes, we propose a latent class model with a focus on the joint analysis of longitudinal and survival processes. Latent class models can help to discover underlying subtypes of subjects in the population, which cannot be observed or defined in advance. Joint models can be used to study the association between longitudinal and survival outcomes. In this dissertation, background knowledge of latent class analysis and jointmodelingmethodsareintroduced. Wethenproposealatentclassmodelforthe joint analysis of longitudinal and survival data, along with details on its likelihood and estimation methods. Simulation studies are provided and asymptotic properties of the estimators are investigated. In the end of the thesis, we applied the proposed model to Terry Beirn Community Programs for Clinical Research on AIDS studies (CPCRA) to characterize the underlying heterogeneity of the cohort and to study the relation between longitudinal CD4 measures and time to death. In summary, our model is desirable when the heterogeneity of subjects cannot be ignored and both the longitudinal and survival outcomes are of interest.

Note: Abstract extracted from PDF text

Degree:
PHD (Doctor of Philosophy)
Language:
English
Rights:
All rights reserved (no additional license for public reuse)
Issued Date:
2013/05/01