Comparison of several independent population means when their samples contain log-normal and possibly zero observations

Xiao Hua Zhou, Wanzhu Tu

Research output: Contribution to journalArticle

56 Citations (Scopus)

Abstract

In this paper, we consider the problem of testing the mean equality of several independent populations that contain log-normal and possibly zero observations. We first showed that the currently used methods in statistical practice, including the nonparametric Kruskal-Wallis test, the standard ANOVA F-test and its two modified versions, the Welch test and the Brown-Forsythe test, could have poor Type I error control. Then we propose a likelihood ratio test that is shown to have much better Type I error control than the existing methods. Finally, we analyze two real data sets that motivated our study using the proposed test.

Original languageEnglish
Pages (from-to)645-651
Number of pages7
JournalBiometrics
Volume55
Issue number2
StatePublished - Jun 1999

Fingerprint

Type I error
Error Control
Zero
Analysis of variance (ANOVA)
Population
Analysis of Variance
testing
sampling
F Test
Testing
Likelihood Ratio Test
Equality
Observation
statistical analysis
analysis of variance
Datasets
Standards
methodology

Keywords

  • Cost data
  • Log-normal
  • Maximum likelihood
  • Skewed distribution
  • Zero costs

ASJC Scopus subject areas

  • Agricultural and Biological Sciences(all)
  • Public Health, Environmental and Occupational Health
  • Agricultural and Biological Sciences (miscellaneous)
  • Applied Mathematics
  • Statistics and Probability

Cite this

Comparison of several independent population means when their samples contain log-normal and possibly zero observations. / Zhou, Xiao Hua; Tu, Wanzhu.

In: Biometrics, Vol. 55, No. 2, 06.1999, p. 645-651.

Research output: Contribution to journalArticle

@article{5d002281dd334b19b32ec741d7ea6b03,
title = "Comparison of several independent population means when their samples contain log-normal and possibly zero observations",
abstract = "In this paper, we consider the problem of testing the mean equality of several independent populations that contain log-normal and possibly zero observations. We first showed that the currently used methods in statistical practice, including the nonparametric Kruskal-Wallis test, the standard ANOVA F-test and its two modified versions, the Welch test and the Brown-Forsythe test, could have poor Type I error control. Then we propose a likelihood ratio test that is shown to have much better Type I error control than the existing methods. Finally, we analyze two real data sets that motivated our study using the proposed test.",
keywords = "Cost data, Log-normal, Maximum likelihood, Skewed distribution, Zero costs",
author = "Zhou, {Xiao Hua} and Wanzhu Tu",
year = "1999",
month = "6",
language = "English",
volume = "55",
pages = "645--651",
journal = "Biometrics",
issn = "0006-341X",
publisher = "Wiley-Blackwell",
number = "2",

}

TY - JOUR

T1 - Comparison of several independent population means when their samples contain log-normal and possibly zero observations

AU - Zhou, Xiao Hua

AU - Tu, Wanzhu

PY - 1999/6

Y1 - 1999/6

N2 - In this paper, we consider the problem of testing the mean equality of several independent populations that contain log-normal and possibly zero observations. We first showed that the currently used methods in statistical practice, including the nonparametric Kruskal-Wallis test, the standard ANOVA F-test and its two modified versions, the Welch test and the Brown-Forsythe test, could have poor Type I error control. Then we propose a likelihood ratio test that is shown to have much better Type I error control than the existing methods. Finally, we analyze two real data sets that motivated our study using the proposed test.

AB - In this paper, we consider the problem of testing the mean equality of several independent populations that contain log-normal and possibly zero observations. We first showed that the currently used methods in statistical practice, including the nonparametric Kruskal-Wallis test, the standard ANOVA F-test and its two modified versions, the Welch test and the Brown-Forsythe test, could have poor Type I error control. Then we propose a likelihood ratio test that is shown to have much better Type I error control than the existing methods. Finally, we analyze two real data sets that motivated our study using the proposed test.

KW - Cost data

KW - Log-normal

KW - Maximum likelihood

KW - Skewed distribution

KW - Zero costs

UR - http://www.scopus.com/inward/record.url?scp=0033044752&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033044752&partnerID=8YFLogxK

M3 - Article

VL - 55

SP - 645

EP - 651

JO - Biometrics

JF - Biometrics

SN - 0006-341X

IS - 2

ER -