簡易檢索 / 詳目顯示
| 研究生： |
蔡騏有 Tsai, Chi-Yu |
|---|---|
| 論文名稱： |
多重檢定中在檢定統計量相關下估計虛無假設為真的個數 A Nonparametric Approach to Estimate the Number of True Null Hypotheses in Multiple Testing under Dependency |
| 指導教授： |
馬瀰嘉
Ma, Mi-Chia |
| 學位類別： |
碩士 Master |
| 系所名稱： |
管理學院 - 統計學系 Department of Statistics |
| 論文出版年： | 2011 |
| 畢業學年度： | 99 |
| 語文別： | 英文 |
| 論文頁數： | 51 |
| 中文關鍵詞： | 整體錯誤率(FWER) 、錯誤拒絕率(FDR) 、白血病 |
| 外文關鍵詞： | Leukemia, Familywise error rate (FWER), False discovery rate (FDR) |
| 相關次數： | 點閱：137 下載：1 |
| 分享至: |
| 查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
在同時檢定多個假設的問題上，控制整體的型一誤差一直是進行多個統計檢定時很重要的議題。有許多相對應的檢定準則可以控制整體錯誤率（the familywise error rate；FWER）或錯誤拒絕率（the false discovery rate；FDR），但當多個假設的某些假設不為真時，控制FWER或FDR的方法都會趨於保守而且檢定力較小，欲改進多個統計檢定的檢定力，方法之一為估計虛無假設為真的個數。本文利用Friedman檢定方法，在檢定統計量相關的假設下估計虛無假設為真的個數並利用統計模擬方法計以標準誤（standard deviation）、均方誤差（mean square error；MSE）來評估所提出的方法表現。最後利用一組白血病（leukemia）患者基因表現資料舉例說明此方法的應用。
In order to handle multiple hypotheses testing﹐it has often been considered as natural to extend the usual Type I error rate﹐it is now well known as the familywise error rate (FWER) or the false discovery rate (FDR). When some hypotheses are not true﹐the FWER-controlling procedures tend to be more conservative and less power in testing. Also﹐estimation of the number of truly null hypotheses﹐which is a key parameter in BH procedure（FDR-controlled）. Thus﹐to estimate the number of true null hypotheses in multiple testing is important in order to control the overall type I error rate and simultaneously to improve the power. A nonparametric approach based on the Friedman test is presented. A simulation study is conducted to evaluate the performances of proposed procedure estimating the number of true null hypotheses by standard deviations﹐and the root mean errors of the estimations. Gene expression data for leukemia patients are used to illustrate the method.
1. Benjamini﹐Y.﹐Hochberg﹐Y. (1995). “Controlling the false discovery rate：a practical and powerful approach to multiple testing”﹐Journal of the Royal Statistical Society. B 57：289-300.
2. Benjamini﹐Y.﹐Hochberg﹐Y. (2000). “On the adaptive control of the false discovery rate in multiple testing with independent statistics”﹐Journal of Educational and Behavioral Statistics. 25：60-83.
3. Cox﹐D. R. (1977). “The role of significance tests”﹐Scandinavian Journal of Statistics. 4﹐49-70.
4. Friguet﹐C.﹐Kloareg﹐M.﹐Causeur﹐D. (2009). “A factor model approach to multiple testing under dependence”﹐Journal of the American Statistical Association﹐104：1406-1415.
5. Golub﹐T. R.﹐Slonim﹐D. K.﹐Tamayo﹐P.﹐Huard﹐C.﹐Gaasenbeek﹐M.﹐Mesirov﹐J. P.﹐Coller﹐H.﹐Loh﹐M. L.﹐Downing﹐J. R.﹐Caligiuri﹐M. A.﹐Bloomfield﹐C. D.﹐Lander E. S. (1999). “Molecular classification of cancer: class discovery and class prediction by gene expression monitoring”﹐Science 286：531-537.
6. Hochberg﹐Y.﹐Tamhane﹐A. C. (1987). Multiple Comparison Procedures﹐New York. John Wiley & Sons.
7. Hsueh﹐H. M.﹐Chen﹐J. J.﹐and Kodell﹐R. L.(2003). “Comparison of methods for estimating the number of true null hypotheses in multiplicity testing”﹐Journal of Biopharmaceutical Statistics. 13﹐675-689.
8. Ma﹐M. C.﹐Chao﹐W. C. (2005). “A nonparametric approach to estimate the number of true null hypotheses in multiple testing”﹐Master essay of Department of Statistics, NCKU.
9. Ma﹐M. C.﹐Wang﹐K. C. (2006). “The estimation for the number of true null hypotheses in multiple testing under dependency”﹐Master essay of Department of Statistics, NCKU.
10. Schweder﹐T.﹐Spjψtvoll﹐E. (1982). “Plots of p-values to evaluate many tests simultaneously”﹐Biometrika 69：493-502.
11. Westfall﹐P. H.﹐Young﹐S. S. (1993). Resampling-Based Multiple Testing﹐New York John Wiley & Sons.
校內：2021-08-28公開校外：2021-08-28公開