Title: 列聯表參數的估計
On the Estimation of Contingency Table Cell Probabilities
Authors: 吳宗益
Wu, Tsung-Yi
周幼珍
Jou Yow-Jen
統計學研究所
Keywords: 初步檢定估計量;收縮估計量;pretest estimator;shrinkage estimator
Issue Date: 1996
Abstract: 本文的動機是要合理的估計一個資料不完整的二維列聯表中各格(cell)的
機率。 我們 的主要想法來自於~ Stein(1956)~年提出的收縮估計量。
我們先給定一個 $\theta_ 0$,藉由檢定$\theta_0$的合理性,獲
得一個初步檢定估計量,由於 此一初步檢定 估計量的值只決定
於$H_0$的成立與否,與檢定統計量$X^2$值的大小 沒有太大關聯。為
了改善此一缺點,而得到了一個較為平滑的收縮估計量。本文的 主要焦
點是比較最大概 似估計量(maximum likelihood estimator)、初步檢定
估計量(pretest estimator)、收 縮估計量(shrinkage estimator)三者
的漸近分配及三者在虛無假設$H_0$、對立假設$H_1 $及~Pitman~對立假
設$K_n$下的漸近風險。
The goal of this thesis is to estimate the cell
probabilities in tw o-way contingency table,with
sampling zeroes.~Our main idea comes from Stei n's shrinkage
estimator.~We propose two estimators for this problem: (1) a
pretest estimator(PTE) for P when some information shows that p
is approxim ately θ,(2)shrinkage estimator is a smoother
version of PTE .~Asymptotic d istribution under the null
hypothesis, a fixed alternativehypothesis and Pitma n-type
alternative hypothesis are derived for both estimators.
~Asymptotic risks are also found and compared with that of MLE
under different hypothes
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT850337001
http://hdl.handle.net/11536/61727
Appears in Collections:Thesis