JSM 2012 Home

JSM 2012 Online Program

The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.

Online Program Home

Abstract Details

Activity Number: 321
Type: Topic Contributed
Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
Sponsor: Section on Bayesian Statistical Science
Abstract - #304260
Title: Learning and Information in Bayesian Joint Models for Longitudinal and Survival Data
Author(s): Laura A. Hatfield*+ and James Hodges and Bradley P Carlin
Companies: Harvard Medical School and University of Minnesota and University of Minnesota
Address: Department of Health Care Policy, Boston, MA, 02115,
Keywords: survival ; Bayesian ; longitudinal ; joint models ; learning
Abstract:

To study a process evolving in time, we may repeatedly quantify features (longitudinal data) and mark the time until an event occurs (survival data). For example, a clinical trial may measure symptom severity and time until death. A broad class of joint models uses latent variables to link longitudinal and survival submodels and analyze these data simultaneously. These models can accommodate many data complexities, yet their properties and performance are not well understood. To quantify the benefit of joint versus separate modeling, we derive posterior means and variances in a simplified normal-lognormal joint model. We study three elements that influence the benefit of joint modeling. First, we show that as the prior variances on fixed effects increase, the benefit of joint modeling under informative priors disappears. Second, because a single latent variable links the two submodels, a coefficient must fix the magnitude and direction of its contribution to one of the submodels. The posterior of this coefficient can exhibit weak identification of its sign when the sample size is small. Finally, we use simulation to study the effect of survival censoring.


The address information is for the authors that have a + after their name.
Authors who are presenting talks have a * after their name.

Back to the full JSM 2012 program




2012 JSM Online Program Home

For information, contact jsm@amstat.org or phone (888) 231-3473.

If you have questions about the Continuing Education program, please contact the Education Department.