Comparison of F test and Resampling Approach for Type I Error Rate and Test Power by Simulation Method


Koskan O., Gurbuz F.

JOURNAL OF AGRICULTURAL SCIENCES-TARIM BILIMLERI DERGISI, vol.15, no.1, pp.105-111, 2009 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 15 Issue: 1
  • Publication Date: 2009
  • Title of Journal : JOURNAL OF AGRICULTURAL SCIENCES-TARIM BILIMLERI DERGISI
  • Page Numbers: pp.105-111

Abstract

In this study, F tests and resampling approaches were compared by simulation method in terms of type I error rate and power of test. The experiments containing 3 groups were chosen. Each groups included equal observations (n=5, 15 and 30) and were drawn from populations. The proportions of variance in samples were found as 1, 5, 10, 30. In order to determine power of test 0.5, 1.0 and 1.5 were added to the values in last group. In the experiments, the samples each including equal observations were drawn from triple combination Z (0, 1), Student t(10), chi(2) (3) and beta (2.5, 5) distributions. Type I error rate and power of the test were obtained from the samples of 10000 simulation trials. Numbers of resampling were determined as 100000 in the experiments. As a result, Resampling tests provided more reliable results than F test for both small sample size and heterogeneous variances.