進階搜尋


 
系統識別號 U0026-0812200910460946
論文名稱(中文) 利用FDR就共變數加以調整之多重假設檢定
論文名稱(英文) The use of FDR in multiple hypotheses testing with covariates adjusted
校院名稱 成功大學
系所名稱(中) 統計學系碩博士班
系所名稱(英) Department of Statistics
學年度 91
學期 2
出版年 92
研究生(中文) 劉怡君
研究生(英文) Yi-Chiung Liu
電子信箱 yjliou0729@yahoo.com.tw
學號 r2690107
學位類別 碩士
語文別 中文
論文頁數 38頁
口試委員 口試委員-李隆安
口試委員-戴政
指導教授-詹世煌
中文關鍵字 多元假設檢定  familywise誤差率  檢定力  Bonferroni多元比較法  偽陽率  逐次P值法  邏輯斯迴歸 
英文關鍵字 Multiple hypotheses testing  Familywise error rate  Power  Bonferroni multiple comparisons method  False discovery rate  P-value  Logistic Regression. 
學科別分類
中文摘要 在許多方面,如遺傳學,我們常需要利用多元假設檢定(multiple hypotheses testing, MHT)以找出具顯著性之因子。例如在人類白血球抗原(human leukocyte antigen, HLA)之一些traits上找出與鼻咽癌有關的對偶基因(alleles)即是。對此種多元假設檢定的情形,以往學者多以控制familywise誤差率(familywise error rate, FWER),如Bonferroni多元比較法,來選取具顯著性之因子,但傳統的Bonferroni法過於保守且其檢定力(power)低。為此Benjamini and Hochberg(1995)和Benjamini and Yekutieli(2001)建議控制偽陽率(false positive rate, FDR),而以逐次P值法(sequential P-value method)來選取顯著因子。逐次P值法於執行上不僅簡便,且相較於FWER控制法(如Bonferroni多元比較法)有較大的檢定力。一般而言,在選取顯著性因子時,多以主要因子為考量依據;如遺傳學上,常著眼於疾病與基因間的關聯,卻忽略了其他因素,譬如年齡、性別等;在本文中,我們將以逐次P值法(sequential P-value method)為基礎,加入其他解釋變數或次要因子予以調整後,再執行多元假設檢定。我們以模擬的方式比較所建議的方法和Benjamini and Hochberg(1995)所提之逐次P值法的績效,發現其檢定力較僅考量單一因子時為高;對所建議的方法,我們以一筆成大醫院病理部所蒐集之一組鼻咽癌資料說明其應用。
英文摘要 Many situations, like genetic research, exist that people need to use mutiple hypotheses testing to search for significant factors related to some kind of event such as disease. One example is to scan over several traits for human leukocyte antigen(HLA)to find out alleles that are related to nasopharyngeal carcinoma. In the case of classical multiple hypotheses testing, people usually select significant factors through the control of familywise error rate (FWER), Bonferroni method is one of such an approach. However, classical methods such as Bonferroni approach suffers loss of power when the number of test hypotheses is large. To overcome this disadvantage, Benjamini and Hochberg(1995)and Benjamini and Yekutieli(2001)suggested, by controlling the false discovery rate (FDR), a sequential p-value method to choose the significant factors. The suggested sequential p-value method is simple to perform, and, as compared to the Bonferroni method which controls the familywise error rate, is more powerful. Generally, we always care about the main factors at selecting the significant factors. In exploring significant factors, some related but not the first line variables such as demographic are also presented. For example, age and sex. In this paper, we, based on the control of FDR, extend the results of Benjamini and Hochberg(1995)and Benjamini and Yekutieli(2001)to the case where covariates involved in the selection of significant factors. We compare our method and the sequential p-value method by simulation. We find that the method we suggested is more powerful. Finally, a microsatellite data sets - NPC data- collected by National Cheng-Kung University Hospital will be analyzed to illustrate the application of our methods.
論文目次 目錄
第一章 緒論 1
1.1 研究動機與背景 1
1.2 研究架構與目的 2
第二章 多元假設檢定 3
2.1 Bonferroni法 3
2.2 False Discovery Rate Procedure 4
2.3 偽陽率FDR(false discovery rate)的定義 5
2.4 逐次P值法(Sequenctial p-value approach) 6
第三章 逐次P值法的修正 8
3.1 修正的原因 8
3.2 共變數對分析的影響 8
第四章 統計模擬 10
4.1 模擬的設定 10
4.1.1 反應變數為連續型 11
4.1.2 反應變數為離散型 15
4.2 分析模式 20
4.3 模擬結果 22
4.3.1 反應變數為連續型之結果 22
4.3.2 反應變數為離散型之結果 28
第五章 實例分析 33
第六章 結論 36
第七章 參考文獻 37
附錄一 38
參考文獻 1. Benjamini, Y. and Hochberg, Y. (1995). Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of Royal Statistical Society, B. 57, 289-300.
2. Benjamini, Y. and Yekutieli, D. (2001). The control of the False Discovery Rate in the Multiple Testing Under Dependency. Annals of Statistics, 29, 1165-1188.
3. Chan, S.H., Day, D.E., Kunaratnam, N., Chia, K.B. and Simons, M.J. (1983) HLA and Nasopharyngeal Carcinoma in Chinese-A Further Study. International Journal of Cancer, 32, 171-176.
4. Lu, C.C., Chen, R.C., Jin, Y.T., Yang, H.B., Chan, S.H. and Tsai, S.T. (2002). Gene Associated with Genetic Susceptibility to Nasopharyngeal Carcinoma is Primary Located within HLA-A Locus in Taiwanese. To appear in International Journal of Cancer.
5. Ooi, E.E., Ren, E.C. and Chan, S.H. (1997). Association Between Microsatellites within the Human MHC and Nasopharyngeal Carcinoma. International Journal of Cancer, 74, 229-232.
6. Ren, E.C., Law, Grace C.T. and Chan, S.H. (1995). HLA-A2 Allelic Microvariants in Nasopharyngeal Carcinoma. International Journal of Cancer, 63, 213-215.
7. Svejgaard, A. and Ryder, L.P. (1994). HLA and Disease Associations: Detecting the Strongest Association. Tissue Antigens, 43, 18-27.
論文全文使用權限
  • 同意授權校內瀏覽/列印電子全文服務,於2003-09-08起公開。
  • 同意授權校外瀏覽/列印電子全文服務,於2003-09-08起公開。


  • 如您有疑問,請聯絡圖書館
    聯絡電話:(06)2757575#65773
    聯絡E-mail:etds@email.ncku.edu.tw