Comparison of ordinary and penalized power-divergence test statistics for small and moderate samples in three-way contingency tables via simulation

Alın A.

COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION, vol.37, no.8, pp.1593-1602, 2008 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 37 Issue: 8
  • Publication Date: 2008
  • Doi Number: 10.1080/03610910802050894
  • Journal Name: COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.1593-1602
  • Keywords: contingency table, log-linear models, penalization, penalized power-divergence measure, power-divergence measure, LOGLINEAR MODELS
  • Dokuz Eylül University Affiliated: Yes

Abstract

In this article, size and power properties of the ordinary and penalized power-divergence test statistics have been compared to test a nested sequence of log-linear models for various penalization and lambda values. 2 x 2 x 2 contingency tables distributed according to a multinomial distribution are considered. The simulation study results reveal that the penalized power-divergence test statistics with penalization value of one and lambda = -1.5, -2 are the best choices for small and moderate samples, respectively.