A Comparison Study in Median Tests

• 발행기관: 한국자료분석학회
• 저자명: 박효일(Hyo-Il Park)
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.17 No.5, 2283~2290쪽, 전체 8쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2015.10.30

영문 초록

In this study, first of all, we review three types of median tests such as the Mood, control and Wilcox, for the two sample problem and then compare the efficiencies of them by obtaining empirical powers through a simulation study. The simulation study will be carried out under the location translation model because all the median tests are known to be effective under the location shift assumption. For this simulation study, we apply the permutation principle to obtain the null distributions for the test statistics with five different distributions for the underlying distributions with varying the sample sizes. Also we derive the asymptotic normality briefly using the large sample approximation theorem and the Bahadur representation theorem for the sample quantile functions. We discuss some interesting features concerning the median tests and related sample medians with the re-sampling methods. Finally we comment about the disadvantage for achieving the nominal significance level for some types of median tests from the simulation results.

목차

1. Introduction

2. Median tests

3. A simulation study

4. Some concluding remarks

References

키워드

Location parameter Median Median test Nonparametric test Permutation principle