決定分析相關性資料時統計檢定力與樣本數的普世強韌法

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/60913

Title:	決定分析相關性資料時統計檢定力與樣本數的普世強韌法
Authors:	羅曉媛;Lo,Hsiao-Yuan
Contributors:	統計研究所
Keywords:	叢集資料;強韌概似函數;多元負二項模型;多元常態模型;cluster data;exchangeable correlation;normal model;negative binomial model;power
Date:	2013-07-17
Issue Date:	2013-08-22 12:06:33 (UTC+8)
Publisher:	國立中央大學
Abstract:	在比較性的研究中，我們可能將叢集資料如家庭成員劃分成對照組與實驗組。Tokola et al. (2011)提出一個不需母體之假設分配的方法來估計對照組與實驗組的差異，此方法亦可估算檢定力。他們的方法主要需正確的假設母體前兩階動差，然而，當動差的假設錯誤時，他們的方法可能也不具有強韌性，分析出來的結果可能是不正確的。本文依據Royall and Tsou (2003)提出的強韌概似函數法，將多元負二項模型及多元常態模型強韌化後探討對照組與實驗組間是否有差異，並與Tokola et al. (2011)的方法比較檢定力與型一誤差機率。 Tokola et al (2011) proposed a statistic for estimating the treatment and control effects difference for cluster data. This statistic requires exchangeable correlation and common variances and is not sensitive to other distributional assumptions. We investigate the validity of their statistic when the second moment assumptions fail. We also robustify the normal and the negative binomial models so that they can adapt themself to the underlying distributions. We compare the three approaches in terms of their power performance.
Appears in Collections:	[Graduate Institute of Statistics] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	610	View/Open

社群 sharing

Loading...