Semiparametric Bayesian Joint Model With Variable Selection

<p>In longitudinal studies, a popular model is the linear mixed model that includes fixed effects and subject specific random effects. In many clinical trials and other medical and reliability studies, we can often obtain repeated measurements or longitudinal data that includes survival or time-to-event histories. Recently, methods for jointly modeling longitudinal and survival data have gained popularity in the statistical literature. In this dissertation, we consider the problem of variable selection in a joint modeling framework where longitudinal and survival data are modeled jointly. Dirichlet process priors are used to relax the parametric assumption of random effects, which has advantages of making the model more robust against possible misspecifications and allows the clustering of subjects. A fully Bayesian method for subset selection of fixed and random effects in joint models is proposed. Simulation examples and an application are used for method evaluation and illustration.</p>

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Campus Access Dissertation

Discipline thesis:degree_discipline

Epidemiology and Biostatistics

Year

2011

Author dc:creator

• Bao, Haikun

Contributors dc:contributor

• Bo Cai

• Biostatistics
• Physical Sciences and Mathematics
• Statistics and Probability

Statement dc:rights

• © 2011, Haikun Bao

Repository record dc:identifier

https://scholarcommons.sc.edu/etd/541

OAI identifier oai:identifier

oai:scholarcommons.sc.edu:etd-1542