簡易檢索 / 詳目顯示

研究生: 許志遠
Hsu, Chih-Yuan
論文名稱: 貝氏逐樣分析之停點研究
A study of stopping time in Bayesian sequential analysis
指導教授: 陳重弘
Chen, Chong-Hong
學位類別: 碩士
Master
系所名稱: 理學院 - 數學系應用數學碩博士班
Department of Mathematics
論文出版年: 2006
畢業學年度: 94
語文別: 中文
論文頁數: 81
中文關鍵詞: 逐樣分析m步前看法終止時間貝氏分析
外文關鍵詞: stopping time, Bayesian analysis, look ahead m-step, sequential analysis
相關次數: 點閱:111下載:2
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
  •   實驗設計的成本,包含決策造成的損失與分析樣本的成本,如何平衡兩者、降低成本,成為我們研究的重點。最佳固定抽樣個數法雖然計算簡易,但無提早停止抽樣的特性,即有固定成本性質。因此,可減少抽樣成本的逐樣分析成為我們關心的對象。逐樣分析是探討在每取完一個觀察值後,決定要繼續抽樣的決策理論。本文中,我們採用貝氏觀點,利用m步前看法進行停點研究,且計算停點的機率分佈。並且由單一母體的簡單隨機抽樣,擴展到分層抽樣的問題。因為逐樣分析的計算繁瑣冗長,所以我利用C++電腦語言幫忙計算出多種情況的停點機率分佈。發現當m步驟愈大,停點的機率分佈有往樣本個數大移的趨勢,即使在分層的抽樣問題上也有相同的結果。

      The overall cost caused by experiment design consists of decision loss and the expenditure of analyzing samples. We will concern about the balance of these two costs and the way to decrease the overall cost. Although the process for calculating the optimal fixed sample size is quite simple, it doesn't stop unless the last sample collected. So it has the property of the fixed overall cost. Therefore, the sequential analysis which may stop early will be concerned by us. The sequential process is to analyze after each sample observed and to decide sampling next observation or not. In the thesis, we use Bayesian viewpoint to proceed with the research of stopping time by look ahead m-step procedure method, and calculate the probabilities of the stopping time at each step. Moreover, I extend the sampling problems from one population to two or more subpopulation. Since the calculating process of the probabilities is tedious, the program written in the language C++ will be provided. As m becomes larger, the probabilities will be distributed to later steps.

    第一章:前言…………………………………………5-6 第二章:背景…………………………………………7-11 第三章:最佳固定樣本個數法 3-1 基本理論………………………………………12-14 3-2 演算法…………………………………………14-16 第四章:逐樣分析法 4-1 一步前看法的步驟……………………………18-27 4-1-1 基本理論……………………………………18-23 4-1-2 終止時間……………………………………23-25 4-2 m步前看法的步驟……………………………27-36 4-2-1 基本理論……………………………………27-29 4-2-2 終止時間……………………………………29-33 4-2-3 演算法………………………………………33-33 4-3 貝氏利益………………………………………37-40 4-3-1 基本理論……………………………………37-38 4-3-2 演算法………………………………………38-40 第五章:分層抽樣的問題 5-1 二層的抽樣……………………………………41-53 5-1-1 最佳固定樣本數……………………………46-47 5-1-2 逐樣分析……………………………………47-50 5-1-3 演算法………………………………………51-51 5-2 一般化…………………………………………53-56 5-2-1 最佳固定樣本數……………………………53-54 5-2-2 逐樣分析……………………………………54-56 第六章:結論 ………………………………………57-57 參考資料 ……………………………………………58-58 附錄:程式 …………………………………………59-81 附錄1 最佳固定樣本個數 ………………………59-62 附錄2 Pr{N=n},m步前看法的步驟(單層)63-67 附錄3 貝氏利益的近似 …………………………68-74 附錄4 Pr{N=n},m步前看法的步驟(二層)75-81

    1.Abdel-Hameed, M. Optimality of the one-step look-ahead stopping times. J. Appl. Probab., Vol.14, P.162-169, 1977.
    2.Berger, James O. Statistical Decision Theory and Bayesian Analysis. Second Edition. Springer-Verlag, New York, 1985.
    3.Cabilio, P. Sequential estimation in Bernoulli trials. Ann Statist., Vol.5, P.342 -356, 1977.
    4.DeGroot, Morris H. Optimal Statistical Decisions. McGraw-Hill, New York, 1970.
    5.Manski, Charles F. Statistical Treatment Rules For Heterogenous Populations. Econometrica, Vol.72, N0.4, P.1221-1246, 2004.
    6.Pham-Gia, T. Distribution of The Stopping Time In Bayesian Sequential Sampling. Austral. & New Zealand J. Statist., Vol.40 (2), P.221-227, 1998.
    7.Pham-Gia, T. Value of the beta prior information. Comm. Statist. A-Theory Methods, Vol.23, P.2175-2195, 1994.
    8.Pham-Gia, T. & Turkkan, N. Sample size determination in Bayesian analysis. The Statistician, Vol.41, P.389-398, 1992.
    9.Wald, A. Statistical Decision Functions. John Wiley & Sons, Inc., New York, 1950.

    下載圖示 校內:2007-07-10公開
    校外:2007-07-10公開
    QR CODE