Department of Statistics
Texas A&M University
STATISTICS COLLOQUIUM
DEPARTMENT OF STATISTICS
and
CENTER FOR ENVIRONMENTAL AND RURAL HEALTH STATISTICS
Texas A&M University
Donglin Zeng
Department of Statistics
University of Michigan
Adjusting for Dependent Censoring by Using
High Dimensional Ausilary Information
ABSTRACT: In survival analysis, people are often interested in estimating the marginal survival function or the effect of a treatment on subject's lifetime using right censored data. Occasionally these estimands may be conditional on a small number of covariates such as race and gender. Suppose the censoring is dependent. But there is sufficient additional auxiliary information that can be used to adjust for the dependent censoring. Then the researcher may choose to model the conditional distribution given both the auxiliary information and treatment. To form the overall treatment effect, the researcher averages the estimated conditional distribution over the distribution of the auxiliary information. This approach is problematic for two reasons. First since this auxiliary information is often high dimensional, it will be easy to misspecify the conditional distribution. And more importantly, misspecification of this conditional distribution (this is a nuisance parameter) will lead to bias in the estimation of the original parameter of interest, the treatment effect parameter. In this talk, I propose the use of two tentative models, the first a tentative model of the survival time given all covariates (both auxiliary information and treatment indicator, race, gender) and the second a tentative model of the censoring time given all covariates. The estimator of the survival function and the estimator of the treatment effect can be obtained by maximizing a pseudo-likelihood that is based on information from the tentative models. I show that the estimators possess double robustness and local efficiency: if one of the two tentative models is correct, they are consistent and are asymptotically normal; if both are correct, they are efficient. Simulation results verify the above double robustness of the estimators.
| DATE: | Thursday, February 22, 2001 | |
| TIME: | 4:00 p.m.-5:00 p.m. | |
| PLACE: | Room 150, Blocker |
Refreshments will be served in the Blocker Building, Room 447, at 3:30 p.m.
Copyright © Unless otherwise stated, all contents of these web pages are copyright by their respective authors or by the Department of Statistics, Texas A&M University, 1995-2001. Comments and questions about this site should be sent to web@stat.tamu.edu.