View : 35 Download: 0

ON SOME ROBUST ALTERNATIVES VIA WINSORIZATION

Title
ON SOME ROBUST ALTERNATIVES VIA WINSORIZATION
Authors
이은경
Issue Date
1995
Department/Major
대학원 통계학과
Keywords
ROBUSTALTERNATIVESVIAWINSORIZATION
Publisher
이화여자대학교 대학원
Degree
Master
Abstract
Most typical methodologies of statistical parametric inferences often require an assumption that observations be drawn from a normal distribution. Because of this very stringent assumption, these methodologies behave badly when the practical situation departs from the ideal assumption, namely, in the presence of outliers. In this regard, many robust inference techniques which are insensitive to underlying assumptions have been developed in order to guard against the possible presence of outliers. In this thesis, we evaluate the values of the power function under the null hypothesis, when H_(0) : δ^(2)=δ^(2)_(0) vs H_(1) : δ^(2)≠δ^(2)_(0) are tested for various situations allowing outliers, and compare the values using the Winsorized variance to that using the adjusted Winsorized variance. And we also evaluate the power functions of the Winsorized sum of squared deviations when H_(0) : δ^(2)=δ^(2)_(0) vs H_(1) : δ^(2)≠δ^(2)_(0) are tested under normality and compare the power of the corrected sum of squares and that of the Winsorized sum of squares. ;대부분의 통계추론방법은 모집단이 정규분포를 따른다는 가정이 필요하다. 그러나, 모집단이 정규본포보다 꼬리가 더 긴 분포를 따르는 경우는 이 추론방법들은 취약성을 갖는다. 이런점에서 가정에 민감하지 않은 많은 로버스트한 추론방법들이 제안되었다. 본 논문에서는 모분산에 대한 로버스트 추정량인 윈저화 분산과 보정된 윈저화 분산을 몬테칼로 모의 실험기법을 이용하여 비교하고 단일표본에서의 분산 검정에서 표본분산을 이용한 경우와 윈저화 분산을 이용한 경우의 검정력 함수를 비교한다.
Fulltext
Show the fulltext
Appears in Collections:
일반대학원 > 통계학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE