|
Activity Number:
|
168
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #305166 |
|
Title:
|
A Generalized Self-Consistency Approach for Joint Modeling Survival and Binary Data
|
|
Author(s):
|
Chen Hu*+ and Alexander Tsodikov
|
|
Companies:
|
University of Michigan and University of Michigan
|
|
Address:
|
1420 Washington Heights, Department of Biostatistics, Ann Arbor, MI, 48109,
|
|
Keywords:
|
Cancer incidence ; Joint Modeling ; Generalized self-consistency ; EM algorithm ; SEER ; Survival and Binary Data
|
|
Abstract:
|
Survival and binary data which are jointly observed and correlated require special attentions in modeling. For example, stage-specific cancer incidence represents a random vector of joint bivariate response represented by the age at diagnosis, and cancer stage. How to link the observed stage-specific cancer incidence with unobserved tumor progression and history of metastasis before diagnosis are of particular interest. In this paper, the joint response is modeled through a series of semiparametric models with time-dependent covariates. We extend the framework of the generalized self-consistency approach (Tsodikov 2003 JRSSB) and use EM algorithm for maximum likelihood estimation. The asymptotic properties of model parameters are derived. This method is illustrated by simulation studies and prostate cancer data from the Surveillance, Epidemiology and End Results (SEER) program.
|
