Dimension-reduced nonparametric maximum likelihood computation for interval-censored data

Authors:
Yong Wang
Affiliations:
Department of Statistics, The University of Auckland, Private Bag 92019, Auckland 1142, New Zealand
Venue:
Computational Statistics & Data Analysis
Year:
2008

Citing 3
Cited 5

Active set algorithms for isotonic regression: a unifying framework

Mathematical Programming: Series A and B
Information bounds and nonparametric maximum likelihood estimation

Information bounds and nonparametric maximum likelihood estimation
Generalized MLE of a joint distribution function with multivariate interval-censored data

Journal of Multivariate Analysis

On testing alternative classes of life distribution with guaranteed survival times

Computational Statistics & Data Analysis
Maximum likelihood computation for fitting semiparametric mixture models

Statistics and Computing
D-optimal designs via a cocktail algorithm

Statistics and Computing
On the interrelation between Dempster-Shafer Belief Structures and their Belief Cumulative Distribution Functions

Knowledge-Based Systems
Efficient computation of nonparametric survival functions via a hierarchical mixture formulation

Statistics and Computing

Quantified Score

Hi-index	0.03

Visualization

Abstract

A general technique is proposed for efficient computation of the nonparametric maximum likelihood estimate (NPMLE) of a survival function. The main idea is to include a new support interval that has the largest gradient value between inclusively every two neighbouring support intervals in the support set at each iteration. It is thus able to expand the support set exponentially fast during the initial stage of computation and tends to produce the same support set of the NPMLE afterward. The use of the proposed technique needs to be combined with an algorithm that can effectively find and remove redundant support intervals, for example, the constrained Newton method, the iterative convex minorant algorithm and the subspace-based Newton method. Numerical studies show that the dimension-reducing technique works very well, especially for purely interval-censored data, where a significant computational improvement via dimension reduction is possible. Strengths and weaknesses of various algorithms are also discussed and demonstrated.