Skip to main navigation Skip to search Skip to main content

Modification of Bartlett's statistic

  • Huimin Fu*
  • , Fengchun Lin
  • *Corresponding author for this work
  • Beihang University

Research output: Contribution to journalArticlepeer-review

Abstract

Bartlett's statistic is widely used in test for homogeneity of variances. It is well known that Bartlett's statistic is an approximate χ2 distribution. However, Bartlett's statistic departs far from its χ2 distribution and results in low test precision in small sample situation. Based on mean equality of both Bartlett's statistic and its χ2 distribution, a new modified formulation of Bartlett's statistic is proposed. It can not only be well distributed as χ2 distribution in both large and small sample situation, but also effectively increase test precision. Comparison between modified and traditional Bartlett's statistic is also presented. Furthermore, test methods for simulation results are established, including standard deviation ratio test for two normal populations, mean ratio test for two exponent populations, and two-step simulation method of standard deviation ratio test for multiple populations.

Original languageEnglish
Pages (from-to)237-240
Number of pages4
JournalJixie Qiangdu/Journal of Mechanical Strength
Volume29
Issue number2
StatePublished - Apr 2007

Keywords

  • Bartlett's statistic
  • Homogeneity of variances
  • Modified Bartlett's statistic
  • Simulation
  • Small sample
  • Test

Fingerprint

Dive into the research topics of 'Modification of Bartlett's statistic'. Together they form a unique fingerprint.

Cite this