Comparing Measures of Location When the Underlying Distribution Has Heavier Tails than Normal

---

Copy For Citation

Özdemir A. F.

Journal of Statisticians: Statistics and Actuarial Sciences, vol.3, pp.8-16, 2010 (Peer-Reviewed Journal)

• Publication Type: Article / Article
• Volume: 3
• Publication Date: 2010
• Journal Name: Journal of Statisticians: Statistics and Actuarial Sciences
• Journal Indexes: TR DİZİN (ULAKBİM)
• Page Numbers: pp.8-16
• Dokuz Eylül University Affiliated: Yes

Abstract

In this study, two conventional (mean and median) and three robust (20% trimmed mean, one-step M-estimator, and modified one-step M-estimator) measures of location are compared in terms of their asymptotic relative efficiencies and mean squared error when the underlying distribution is contaminated normal. When n=20, one-step M-estimator was best in five, modified one-step M-estimator was best in three, and 20% trimmed mean was best in one situation, when n=40, one-step M-estimator and modified one-step M-estimator were best in four, 20% trimmed mean was best in one sampling situation covered.