翻訳と辞書
Words near each other
・ Proof of Existence
・ Proof of Fermat's Last Theorem
・ Proof of Fermat's Last Theorem for specific exponents
・ Proof of funds
・ Proof of Impact
・ Proof of impossibility
・ Proof of insurance
・ Proof of knowledge
・ Proof of Life
・ Proof of Life (album)
・ Proof of Life (disambiguation)
・ Proof of Life (Scott Stapp album)
・ Proof of Life (The Bill)
・ Proof of O(log*n) time complexity of union–find
・ Proof of purchase
Proof of Stein's example
・ Proof of the Euler product formula for the Riemann zeta function
・ Proof of the Man
・ Proof of Youth
・ Proof positive
・ Proof Positive (album)
・ Proof Positive (Greene story)
・ Proof Positive (TV series)
・ Proof procedure
・ Proof School
・ Proof sketch for Gödel's first incompleteness theorem
・ Proof test
・ Proof that 22/7 exceeds π
・ Proof that e is irrational
・ Proof That the Youth Are Revolting


Dictionary Lists
翻訳と辞書 辞書検索 [ 開発暫定版 ]
スポンサード リンク

Proof of Stein's example : ウィキペディア英語版
Proof of Stein's example

Stein's example is an important result in decision theory which can be stated as
: ''The ordinary decision rule for estimating the mean of a multivariate Gaussian distribution is inadmissible under mean squared error risk in dimension at least 3''.
The following is an outline of its proof. The reader is referred to the main article for more information.
==Sketched proof==
The risk function of the decision rule d(\mathbf) = \mathbf is
:R(\theta,d) = \mathbb_\theta(|\mathbf|^2 )
::=\int (\mathbf)^T(\mathbf) \left( \frac \right)^ e^) } m(dx)
:: = n.\,
Now consider the decision rule
:d'(\mathbf) = \mathbf - \frac\mathbf
where \alpha = n-2. We will show that d' is a better decision rule than d. The risk function is
:R(\theta,d') = \mathbb_\theta\left(\left|\mathbf + \frac\mathbf\right|^2\right )
:: = \mathbb_\theta\left(|\mathbf|^2 + 2(\mathbf)^T\frac\mathbf + \frac|\mathbf|^2 \right )
:: = \mathbb_\theta\left(|\mathbf|^2 \right ) + 2\alpha\mathbb_\theta\left(\right )
— a quadratic in \alpha. We may simplify the middle term by considering a general "well-behaved" function h:\mathbf \mapsto h(\mathbf) \in \mathbb and using integration by parts. For 1\leq i \leq n, for any continuously differentiable h growing sufficiently slowly for large x_i we have:
::\mathbb_\theta ((\theta_i - X_i) h(\mathbf) | X_j=x_j (j\neq i) )= \int (\theta_i - x_i) h(\mathbf) \left( \frac \right)^ e^ } m(dx_i)
:= \left(h(\mathbf) \left( \frac \right)^ e^ } \right )^\infty_
- \int \frac(\mathbf) \left( \frac \right)^ e^ } m(dx_i)
: = - \mathbb_\theta \left(\frac(\mathbf) | X_j=x_j (j\neq i) \right ).

Therefore,
:\mathbb_\theta ((\theta_i - X_i) h(\mathbf) )= - \mathbb_\theta \left(\frac(\mathbf) \right ).
(This result is known as Stein's lemma.)
Now, we choose
:
h(\mathbf) = \frac.

If h met the "well-behaved" condition (it doesn't, but this can be remedied -- see below), we would have
:\frac = \frac - \frac
and so
::
\mathbb_\theta\left((\theta_i - X_i) \frac \right )
: = - \sum_^n \mathbb_\theta \left(\frac - \frac \right )
: = -(n-2)\mathbb_\theta \left().
Then returning to the risk function of d' :
:
R(\theta,d') = n - 2\alpha(n-2)\mathbb_\theta\left() + \alpha^2\mathbb_\theta\left(\right ).

This quadratic in \alpha is minimized at
:\alpha = n-2,\,
giving
:R(\theta,d') = R(\theta,d) - (n-2)^2\mathbb_\theta\left(\right )
which of course satisfies:
:
R(\theta,d') < R(\theta,d).

making d an inadmissible decision rule.
It remains to justify the use of
:
h(\mathbf)= \frac|^2}.

This function is not continuously differentiable since it is singular at \mathbf=0. However the function
:
h(\mathbf) = \frac|^2}

is continuously differentiable, and after following the algebra through and letting \epsilon \to 0 one obtains the same result.


抄文引用元・出典: フリー百科事典『 ウィキペディア(Wikipedia)
ウィキペディアで「Proof of Stein's example」の詳細全文を読む



スポンサード リンク
翻訳と辞書 : 翻訳のためのインターネットリソース

Copyright(C) kotoba.ne.jp 1997-2016. All Rights Reserved.