Activity Number:
|
11
|
Type:
|
Topic Contributed
|
Date/Time:
|
Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
IMS
|
Abstract - #301260 |
Title:
|
On the Identifiability of the Nonparametric Maximum Likelihood Survival Estimator
|
Author(s):
|
Alain Vandal*+ and Robert Gentleman and Xuecheng Liu
|
Affiliation(s):
|
McGill University and Harvard University and McGill University
|
Address:
|
Burnside Hall, 805, rue Sherbrooke ouest, Montréal, Quebec, H3A 2K6, Canada
|
Keywords:
|
censored data ; graph theory ; survival estimation
|
Abstract:
|
The general case of interval censored data encompasses the well-studied cases of current status data, doubly-censored data, and right-censored data. We use very basic graph-theoretical concepts to highlight some fundamental differences between the univariate and the bivariate nonparametric maximum likelihood estimates (NPMLE) of the survival function. In particular, two types of unidentifiabilities may beset the NPMLE. The more fundamental of the two, which we call mixture unidentifiability, can only appear in the multivariate case. We show how mixture unidentifiability is related to the appearance of cycles in the interval graph of the data. We also show how to characterize all likelihood-maximizing solutions by expressing them as a convex combination of optimal solutions. We also discuss the consequences of unidentifiability on some estimators based on the NPMLE.
|
- 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 2002 program |