標題: Trimmed mean是否可以經由Winsorized mean改善它的有效性
Can the Asymptotic Efficiences of the Trimmed Means be Improved by Winsorized Means
作者: 王則堯
陳鄰安
統計學研究所
關鍵字: Winsorized mean;Trimmed mean;Regression quantile;Residual quantile;Symmetric quantile
公開日期: 1999
摘要: 在線性回歸模式中,我們發展出三套Winsorized Means的理論:
(1).Koenker and Bassett(1978) regression quantiles
(2).Chen and Chiang(1996) symmetric quantile
(3).Welsh(1987) residual quantile
在計算有效性的過程中,我們發現下面一些事實:
(1).Winsorized Means不能改善Koenker and Bassett's and Welsh's
Trimmed Means 的有效性.
(2).Winsorized Means不但可以改善symmetric Trimmed Means,
而且它會更接近Cramer-Rao lower bound.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT880337015
http://hdl.handle.net/11536/65382
顯示於類別:畢業論文