Nonparametric inference in factorial designs with censored data

Michael G. Akritas, Michael P. LaValley

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

A method is proposed for testing the hypotheses of no main effects and no interaction in factorial designs with several observations per cell. The method uses the fact that these hypotheses can be expressed in terms of a vector of contrasts. It is based on the observation that nonparametric estimation of these contrasts is no more difficult than estimation of the location difference in the two-sample problem. To implement the method with censored data, a new extension of the Hodges-Lehmann estimator is proposed. The estimator is simple to compute and its variance is easily evaluated. A simulation study examines the performance of the proposed estimation and testing method in the context of a two-by-two design, and a real data set from a three-way layout with heavy censoring is analyzed.

Original languageEnglish (US)
Pages (from-to)913-924
Number of pages12
JournalBiometrics
Volume52
Issue number3
DOIs
StatePublished - Jan 1 1996

Fingerprint

Nonparametric Inference
Factorial Design
Censored Data
Testing
Two-sample Problem
Estimator
Main Effect
Nonparametric Estimation
Censoring
methodology
Layout
testing
Simulation Study
Cell
Interaction
cells
sampling
Observation

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Biochemistry, Genetics and Molecular Biology(all)
  • Immunology and Microbiology(all)
  • Agricultural and Biological Sciences(all)
  • Applied Mathematics

Cite this

Akritas, Michael G. ; LaValley, Michael P. / Nonparametric inference in factorial designs with censored data. In: Biometrics. 1996 ; Vol. 52, No. 3. pp. 913-924.
@article{d5ee330cd3824f48a88eb8179c2b7d8c,
title = "Nonparametric inference in factorial designs with censored data",
abstract = "A method is proposed for testing the hypotheses of no main effects and no interaction in factorial designs with several observations per cell. The method uses the fact that these hypotheses can be expressed in terms of a vector of contrasts. It is based on the observation that nonparametric estimation of these contrasts is no more difficult than estimation of the location difference in the two-sample problem. To implement the method with censored data, a new extension of the Hodges-Lehmann estimator is proposed. The estimator is simple to compute and its variance is easily evaluated. A simulation study examines the performance of the proposed estimation and testing method in the context of a two-by-two design, and a real data set from a three-way layout with heavy censoring is analyzed.",
author = "Akritas, {Michael G.} and LaValley, {Michael P.}",
year = "1996",
month = "1",
day = "1",
doi = "10.2307/2533052",
language = "English (US)",
volume = "52",
pages = "913--924",
journal = "Biometrics",
issn = "0006-341X",
publisher = "Wiley-Blackwell",
number = "3",

}

Nonparametric inference in factorial designs with censored data. / Akritas, Michael G.; LaValley, Michael P.

In: Biometrics, Vol. 52, No. 3, 01.01.1996, p. 913-924.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Nonparametric inference in factorial designs with censored data

AU - Akritas, Michael G.

AU - LaValley, Michael P.

PY - 1996/1/1

Y1 - 1996/1/1

N2 - A method is proposed for testing the hypotheses of no main effects and no interaction in factorial designs with several observations per cell. The method uses the fact that these hypotheses can be expressed in terms of a vector of contrasts. It is based on the observation that nonparametric estimation of these contrasts is no more difficult than estimation of the location difference in the two-sample problem. To implement the method with censored data, a new extension of the Hodges-Lehmann estimator is proposed. The estimator is simple to compute and its variance is easily evaluated. A simulation study examines the performance of the proposed estimation and testing method in the context of a two-by-two design, and a real data set from a three-way layout with heavy censoring is analyzed.

AB - A method is proposed for testing the hypotheses of no main effects and no interaction in factorial designs with several observations per cell. The method uses the fact that these hypotheses can be expressed in terms of a vector of contrasts. It is based on the observation that nonparametric estimation of these contrasts is no more difficult than estimation of the location difference in the two-sample problem. To implement the method with censored data, a new extension of the Hodges-Lehmann estimator is proposed. The estimator is simple to compute and its variance is easily evaluated. A simulation study examines the performance of the proposed estimation and testing method in the context of a two-by-two design, and a real data set from a three-way layout with heavy censoring is analyzed.

UR - http://www.scopus.com/inward/record.url?scp=0029840802&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0029840802&partnerID=8YFLogxK

U2 - 10.2307/2533052

DO - 10.2307/2533052

M3 - Article

C2 - 8805761

AN - SCOPUS:0029840802

VL - 52

SP - 913

EP - 924

JO - Biometrics

JF - Biometrics

SN - 0006-341X

IS - 3

ER -