|
Activity Number:
|
501
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306343 |
|
Title:
|
Proportional Hazards Model with Empirically Estimated Weights
|
|
Author(s):
|
Qing Pan*+ and Douglas E. Schaubel
|
|
Companies:
|
University of Michigan and University of Michigan
|
|
Address:
|
Biostat, 1518 Gilbert Court F21, Ann Arbor, MI, 48105,
|
|
Keywords:
|
survival analysis ; selection bias ; inverse probability weighting ; weighted Cox model ; empirical weights ; logistic regression
|
|
Abstract:
|
In observational studies, the study population is often a biased sample of the underlying target population. Weighted Cox models have been developed for settings in which weights are fixed (e.g., chosen by the investigator). However, in many practical settings, the weights are estimated and must be treated as such for accurate inference. We propose a two-stage procedure: weights are estimated through a parametric model fitted to a representative sample from the target population at the first stage; with a weighted proportional hazards model fitted to the biased sample at the second stage. Estimators for the regression parameter and cumulative baseline hazard are proposed. Asymptotic properties of the parameter estimators are derived, with their applicability to finite samples evaluated through simulation. The proposed method is applied to transplant survival data.
|
