Improved power in multinomial goodness-of-fit tests

doi:10.1111/1467-9884.00325

If you are seeing this message, you may be experiencing temporary network problems. Please wait a few minutes and refresh the page. If the problem persists, you may wish to report it to your local Network Manager.

It is also possible that your web browser is not configured or not able to display style sheets. In this case, although the visual presentation will be degraded, the site should continue to be functional. We recommend using the latest version of Microsoft or Mozilla web browser to help minimise these problems.

Wiley InterScience

Journal of the Royal Statistical Society: Series D (The Statistician)

Volume 51 Issue 3, Pages 381 - 393

Published Online: 9 Aug 2002

2003 Royal Statistical Society

Published on behalf of

Go to Society Site

< Previous Abstract | Next Abstract >

Save Article to My Profile Download Citation Request Permissions

Abstract | References | Full Text: HTML, PDF (Size: 369K) | Related Articles | Citation Tracking

Improved power in multinomial goodness-of-fit tests

Ayanendranath Basu, Surajit Ray, Chanseok Parkand Srabashi Basu

Indian Statistical Institute, Calcutta, IndiaPenn State University, University Park, USAClemson University, USAIndian Statistical Institute, Calcutta, India

Address for Correspondence: Surajit Ray, Department of Statistics, Penn State University, University Park, PA 16802, USA. E-mail: surajit@stat.psu.edu

KEYWORDS

Disparities • Empty cell penalty • Goodness of fit • Power divergence

ABSTRACT

Summary.Pearson's χ²- and the log-likelihood ratio χ²-statistics are fundamental tools in goodness-of-fit testing. Cressie and Read constructed a general family of divergences which includes both statistics as special cases. This family is indexed by a single parameter, and divergences at either end of the scale are more powerful against alternatives of one type while being rather poor against the opposite type. Here we present several new goodness-of-fit testing procedures which have reasonably high powers for both kinds of alternative. Graphical studies illustrate the advantages of the new methods.

[Received March 2001. Revised April 2002]

DIGITAL OBJECT IDENTIFIER (DOI)

10.1111/1467-9884.00325 About DOI