Abstract

For right-censored data perhaps the most commonly used tests are weighted logrank tests, such as the logrank and Wilcoxon-type tests. In this paper we review several generalizations of those weighted logrank tests to interval-censored data and present an R package, interval, to implement many of them. The interval package depends on the perm package, also presented here, which performs exact and asymptotic linear permutation tests. The perm package performs many of the tests included in the already available coin package, and provides an independent validation of coin. We review analysis methods for interval-censored data, and we describe and show how to use the interval and perm packages.

Highlights

  • In their original paper introducing the logrank test, Peto and Peto (1972) outlined how to perform permutation-based logrank tests for interval-censored data

  • In this paper we describe an R package, called interval, to perform weighted logrank tests (WLRT) for intervalcensored data

  • This application section demonstrates the use of two main functions of interval: icfit which provide nonparametric maximum likelihood estimators (NPMLEs) of the survival distribution and ictest which performs weighted logrank tests

Read more

Summary

Introduction

In their original paper introducing the logrank test, Peto and Peto (1972) outlined how to perform permutation-based logrank tests for interval-censored data. We know of no readily available software to perform these tests (or other different generalizations of the WLRTs to intervalcensored data), except for the S-PLUS functions (written by the first author) upon which the interval package is based (Fay 1999b). This application section demonstrates the use of two main functions of interval: icfit which provide nonparametric maximum likelihood estimators (NPMLEs) of the survival distribution and ictest which performs weighted logrank tests. All implementations of the WLRTs require first fitting an NPMLE of the overall survival distribution and, for some kinds of inferences, permutation methods may be used Because of this latter fact, the interval depends on the perm package which performs exact and asymptotic linear permutation tests. We show how the wlr_trafo function of the interval package may be called seamlessly from within coin to perform exact permutation tests on interval-censored data possibly more quickly than using the default perm package

Censoring assumptions
Nonparametric estimation of survival distributions
Overview of weighted logrank tests
Choosing the likelihood for WLR tests
Choosing the inferential method for WLR tests
Regression in interval-censored data
Mathematical formulation of the scores for the WLRT
Application
Two-sample weighted logrank tests
Other test options
The perm package
Design and implementation
Findings
Interacting with the coin package
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call