Efficiencies of Partial-Likelihood-Based Inferences Concerning Survival Regression Models.

Abstract

This paper addresses four important issues in large-simple survival analysis: 1) the inadequacy of a full-liklihood approach to the problem of testing, with right-censored survival data and within a regression model for hazards of failure, the hypothesis that the mechanism of failure does not depend on a designated subset of time-dependent covariates; 2) the definition of an appropriate 'full-likelihood', for survival problems with time-dependent covariate processes Z sub i (t), by means of Cox's (1975) notion of Partial Likelihood; 3) general justification of the assertion by Peto and Peto (1972) of 'asymptotic efficiency' in a semiparametric sense for partial-likelihood-score test-statistics; and 4) formulation of semiparametric efficiency questions in the Survival setting (whether at the null-hypothesis or not), which have been described previously (Begun et al. 1983) as Hilbert-space projection problems, concretely in terms of calculus-of-variations problems. (Author)

Document Details

Document Type
Technical Report
Publication Date
Dec 18, 1986
Accession Number
ADA176771

Entities

People

  • Eric V. Slud

Organizations

  • University of Maryland

Tags

DTIC Thesaurus Topics

  • Calculus
  • Calculus Of Variations
  • Data Science
  • Efficiency
  • Hilbert Space
  • Information Science
  • Mathematics
  • Statistics
  • Survival

Fields of Study

  • Mathematics

Readers

  • Statistical inference.
  • Theoretical Analysis.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms
  • Space