<?xml version="1.0"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
	<id>https://en.formulasearchengine.com/w/api.php?action=feedcontributions&amp;feedformat=atom&amp;user=130.184.0.0%2F16</id>
	<title>formulasearchengine - User contributions [en]</title>
	<link rel="self" type="application/atom+xml" href="https://en.formulasearchengine.com/w/api.php?action=feedcontributions&amp;feedformat=atom&amp;user=130.184.0.0%2F16"/>
	<link rel="alternate" type="text/html" href="https://en.formulasearchengine.com/wiki/Special:Contributions/130.184.0.0/16"/>
	<updated>2026-09-20T01:37:06Z</updated>
	<subtitle>User contributions</subtitle>
	<generator>MediaWiki 1.47.0-wmf.7</generator>
	<entry>
		<id>https://en.formulasearchengine.com/w/index.php?title=59_(number)&amp;diff=4557</id>
		<title>59 (number)</title>
		<link rel="alternate" type="text/html" href="https://en.formulasearchengine.com/w/index.php?title=59_(number)&amp;diff=4557"/>
		<updated>2013-09-13T21:47:58Z</updated>

		<summary type="html">&lt;p&gt;130.184.19.175: /* In sports */&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;{{Unreferenced|date=October 2009}}&lt;br /&gt;
&lt;br /&gt;
In [[statistics]], the &#039;&#039;&#039;Rao–Blackwell theorem&#039;&#039;&#039;, sometimes referred to as the &#039;&#039;&#039;Rao–Blackwell–Kolmogorov theorem&#039;&#039;&#039;, is a result which characterizes the transformation of an arbitrarily crude [[estimator]] into an estimator that is optimal by the [[mean squared error|mean-squared-error]] criterion or any of a variety of similar criteria. &lt;br /&gt;
&lt;br /&gt;
The Rao–Blackwell theorem states that if &#039;&#039;g&#039;&#039;(&#039;&#039;X&#039;&#039;) is any kind of [[estimator]] of a parameter θ, then the [[conditional expectation]] of &#039;&#039;g&#039;&#039;(&#039;&#039;X&#039;&#039;) given &#039;&#039;T&#039;&#039;(&#039;&#039;X&#039;&#039;), where &#039;&#039;T&#039;&#039; is a [[sufficient statistic]], is typically a better estimator of θ, and is never worse. Sometimes one can very easily construct a very crude estimator &#039;&#039;g&#039;&#039;(&#039;&#039;X&#039;&#039;), and then evaluate that conditional expected value to get an estimator that is in various senses optimal.&lt;br /&gt;
&lt;br /&gt;
The theorem is named after [[Calyampudi Radhakrishna Rao]] and [[David Blackwell]].  The process of transforming an estimator using the Rao–Blackwell theorem is sometimes called &#039;&#039;&#039;Rao–Blackwellization&#039;&#039;&#039;. The transformed [[estimator]] is called the &#039;&#039;&#039;Rao–Blackwell estimator&#039;&#039;&#039;.&lt;br /&gt;
&lt;br /&gt;
==Definitions==&lt;br /&gt;
*An [[estimator]] δ(&#039;&#039;X&#039;&#039;) is an &#039;&#039;observable&#039;&#039; random variable (i.e. a [[statistic]]) used for estimating some &#039;&#039;unobservable&#039;&#039; quantity. For example, one may be unable to observe the average height of &#039;&#039;all&#039;&#039; male students at the University of X, but one may observe the heights of a random sample of 40 of them.  The average height of those 40—the &amp;quot;sample average&amp;quot;—may be used as an estimator of the unobservable &amp;quot;population average&amp;quot;.&lt;br /&gt;
&lt;br /&gt;
*A [[sufficiency (statistics)|sufficient statistic]] &#039;&#039;T&#039;&#039;(&#039;&#039;X&#039;&#039;) is a statistic calculated from data &#039;&#039;X&#039;&#039; to estimate some parameter θ for which it is true that no other statistic which can be calculated from data X provides any additional information about θ. It is defined as an &#039;&#039;observable&#039;&#039; [[random variable]] such that the [[conditional probability]] distribution of all observable data &#039;&#039;X&#039;&#039; given &#039;&#039;T&#039;&#039;(&#039;&#039;X&#039;&#039;) does not depend on the &#039;&#039;unobservable&#039;&#039; parameter θ, such as the mean or standard deviation of the whole population from which the data &#039;&#039;X&#039;&#039; was taken. In the most frequently cited examples, the &amp;quot;unobservable&amp;quot; quantities are parameters that parametrize a known family of [[probability distribution]]s according to which the data are distributed.&lt;br /&gt;
::In other words, a [[sufficiency (statistics)|sufficient statistic]] &#039;&#039;T(X)&#039;&#039; for a parameter θ is a [[statistic]] such that the [[conditional probability|conditional distribution]] of the data &#039;&#039;X&#039;&#039;, given &#039;&#039;T&#039;&#039;(&#039;&#039;X&#039;&#039;), does not depend on the parameter θ.&lt;br /&gt;
&lt;br /&gt;
*A &#039;&#039;&#039;Rao–Blackwell estimator&#039;&#039;&#039; δ&amp;lt;sub&amp;gt;1&amp;lt;/sub&amp;gt;(&#039;&#039;X&#039;&#039;) of an unobservable quantity θ is the conditional [[expected value]] E(δ(&#039;&#039;X&#039;&#039;) | &#039;&#039;T&#039;&#039;(&#039;&#039;X&#039;&#039;)) of some estimator δ(&#039;&#039;X&#039;&#039;) given a sufficient statistic &#039;&#039;T&#039;&#039;(&#039;&#039;X&#039;&#039;).  Call δ(&#039;&#039;X&#039;&#039;) the &#039;&#039;&#039;&amp;quot;original estimator&amp;quot;&#039;&#039;&#039; and δ&amp;lt;sub&amp;gt;1&amp;lt;/sub&amp;gt;(&#039;&#039;X&#039;&#039;) the &#039;&#039;&#039;&amp;quot;improved estimator&amp;quot;&#039;&#039;&#039;.  It is important that the improved estimator be &#039;&#039;observable&#039;&#039;, i.e. that it not depend on θ.  Generally, the conditional expected value of one function of these data given another function of these data &#039;&#039;does&#039;&#039; depend on θ, but the very definition of sufficiency given above entails that this one does not.&lt;br /&gt;
&lt;br /&gt;
*The &#039;&#039;[[mean squared error]]&#039;&#039; of an estimator is the expected value of the square of its deviation from the unobservable quantity being estimated.&lt;br /&gt;
&lt;br /&gt;
==The theorem==&lt;br /&gt;
&lt;br /&gt;
===Mean-squared-error version===&lt;br /&gt;
One case of Rao–Blackwell theorem states:&lt;br /&gt;
&lt;br /&gt;
:The mean squared error of the Rao–Blackwell estimator does not exceed that of the original estimator.&lt;br /&gt;
&lt;br /&gt;
In other words&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt;\operatorname{E}((\delta_1(X)-\theta)^2)\leq &lt;br /&gt;
       \operatorname{E}((\delta(X)-\theta)^2).\,\!&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
The essential tools of the proof besides the definition above are the [[law of total expectation]] and the fact that for any random variable &#039;&#039;Y&#039;&#039;, E(&#039;&#039;Y&#039;&#039;&amp;lt;sup&amp;gt;2&amp;lt;/sup&amp;gt;) cannot be less than [E(&#039;&#039;Y&#039;&#039;)]&amp;lt;sup&amp;gt;2&amp;lt;/sup&amp;gt;.  That inequality is a case of [[Jensen&#039;s inequality]], although it may also be shown to follow instantly from the frequently mentioned fact that&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt; 0 \leq \operatorname{Var}(Y) = \operatorname{E}((Y-\operatorname{E}(Y))^2) = &lt;br /&gt;
               \operatorname{E}(Y^2)-(\operatorname{E}(Y))^2.\,\!&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
===Convex loss generalization===&lt;br /&gt;
The more general version of the Rao–Blackwell theorem speaks of the &amp;quot;expected loss&amp;quot; or [[risk function]]:&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt;\operatorname{E}(L(\delta_1(X)))\leq \operatorname{E}(L(\delta(X)))\,\!&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
where the &amp;quot;loss function&amp;quot; &#039;&#039;L&#039;&#039; may be any [[convex function]].  For the proof of the more general version, Jensen&#039;s inequality cannot be dispensed with.&lt;br /&gt;
&lt;br /&gt;
==Properties==&lt;br /&gt;
The improved estimator is [[bias of an estimator|unbiased]] if and only if the original estimator is unbiased, as may be seen at once by using the [[law of total expectation]].  The theorem holds regardless of whether biased or unbiased estimators are used.&lt;br /&gt;
&lt;br /&gt;
The theorem seems very weak: it says only that the Rao–Blackwell estimator is no worse than the original estimator.  In practice, however, the improvement is often enormous.&lt;br /&gt;
&lt;br /&gt;
==Example==&lt;br /&gt;
Phone calls arrive at a switchboard according to a [[Poisson process]] at an average rate of λ per minute.  This rate is not observable, but the numbers &#039;&#039;X&#039;&#039;&amp;lt;sub&amp;gt;1&amp;lt;/sub&amp;gt;, ..., &#039;&#039;X&#039;&#039;&amp;lt;sub&amp;gt;&#039;&#039;n&#039;&#039;&amp;lt;/sub&amp;gt; of phone calls that arrived during &#039;&#039;n&#039;&#039; successive one-minute periods are observed.  It is desired to estimate the probability &#039;&#039;e&#039;&#039;&amp;lt;sup&amp;gt;&amp;amp;minus;λ&amp;lt;/sup&amp;gt; that the next one-minute period passes with no phone calls.  &lt;br /&gt;
&lt;br /&gt;
An &#039;&#039;extremely&#039;&#039; crude estimator of the desired probability is&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt;\delta_0=\left\{\begin{matrix}1 &amp;amp; \text{if}\ X_1=0, \\&lt;br /&gt;
0 &amp;amp; \text{otherwise,}\end{matrix}\right.&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
i.e., it estimates this probability to be 1 if no phone calls arrived in the first minute and zero otherwise.  Despite the apparent limitations of this estimator, the result given by its Rao–Blackwellization is a very good estimator.&lt;br /&gt;
&lt;br /&gt;
The sum&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt; S_n = \sum_{i=1}^n X_{i} = X_1+\cdots+X_n\,\!&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
can be readily shown to be a sufficient statistic for λ, i.e., the &#039;&#039;conditional&#039;&#039; distribution of the data &#039;&#039;X&#039;&#039;&amp;lt;sub&amp;gt;1&amp;lt;/sub&amp;gt;, ..., &#039;&#039;X&#039;&#039;&amp;lt;sub&amp;gt;&#039;&#039;n&#039;&#039;&amp;lt;/sub&amp;gt;, depends on λ only through this sum.  Therefore, we find the Rao–Blackwell estimator&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt;\delta_1=\operatorname{E}(\delta_0\mid S_n=s_n).\,\!&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
After doing some algebra we have&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt;&lt;br /&gt;
\begin{align}&lt;br /&gt;
 \delta_1 &amp;amp; = \operatorname{E}(\operatorname{1}_{\{X_1=0\}} \mid \sum_{i=1}^n X_{i} = s_n)=P(X_{1}=0 \mid \sum_{i=1}^n X_{i} = s_n) \\&lt;br /&gt;
&amp;amp; = \frac{P(X_{1}=0, \sum_{i=2}^n X_{i} = s_n)}{P(\sum_{i=1}^n X_{i} = s_n)} = e^{-\lambda}\frac{\left((n-1)\lambda\right)^{s_n}e^{-(n-1)\lambda}}{s_n!} \Bigg / \frac{(n\lambda)^{s_n}e^{-n\lambda}}{s_n!} = \left(1-{1 \over n}\right)^{s_n}.&lt;br /&gt;
\end{align}&lt;br /&gt;
&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
Since the average number of calls arriving during the first &#039;&#039;n&#039;&#039; minutes is &#039;&#039;n&#039;&#039;λ, one might not be surprised if this estimator has a fairly high probability (if &#039;&#039;n&#039;&#039; is big) of being close to&lt;br /&gt;
&lt;br /&gt;
:&amp;lt;math&amp;gt;\left(1-{1 \over n}\right)^{n\lambda}\approx e^{-\lambda}.&amp;lt;/math&amp;gt;&lt;br /&gt;
&lt;br /&gt;
So δ&amp;lt;sub&amp;gt;1&amp;lt;/sub&amp;gt; is clearly a very much improved estimator of that last quantity.  In fact, since &#039;&#039;S&#039;&#039;&amp;lt;sub&amp;gt;&#039;&#039;n&#039;&#039;&amp;lt;/sub&amp;gt; is [[completeness (statistics)|complete]] and δ&amp;lt;sub&amp;gt;0&amp;lt;/sub&amp;gt; is unbiased, δ&amp;lt;sub&amp;gt;1&amp;lt;/sub&amp;gt; is the unique minimum variance unbiased estimator by the [[Lehmann–Scheffé theorem]].&lt;br /&gt;
&lt;br /&gt;
==Idempotence==&lt;br /&gt;
The Rao–Blackwell process is [[idempotent]].  Using it to improve the already improved estimator does not obtain a further improvement, but merely returns as its output the same improved estimator.&lt;br /&gt;
&lt;br /&gt;
==Completeness and Lehmann&amp;amp;ndash;Scheffé minimum variance==&lt;br /&gt;
If the conditioning statistic is both complete and sufficient, and the starting estimator is unbiased, then the Rao&amp;amp;ndash;Blackwell estimator is the unique &amp;quot;[[minimum-variance unbiased estimator|best unbiased estimator]]&amp;quot;: see [[Lehmann–Scheffé theorem]].&lt;br /&gt;
&lt;br /&gt;
==See also==&lt;br /&gt;
* [[Basu&#039;s theorem]] &amp;amp;mdash; Another result on complete sufficient and ancillary statistics&lt;br /&gt;
&lt;br /&gt;
==References==&lt;br /&gt;
{{Refbegin}}&lt;br /&gt;
* {{Cite journal&lt;br /&gt;
 |last=Blackwell |first=D. |authorlink=David Blackwell&lt;br /&gt;
 |title=Conditional expectation and unbiased sequential estimation&lt;br /&gt;
 |journal=[[Annals of Mathematical Statistics]]&lt;br /&gt;
 |volume=18 |issue=1 |pages=105–110 |year=1947&lt;br /&gt;
 |doi=10.1214/aoms/1177730497 |mr=19903 | zbl = 0033.07603&lt;br /&gt;
}}&lt;br /&gt;
*{{Cite journal&lt;br /&gt;
 |last=Kolmogorov |first=A. N. |authorlink=Andrey Kolmogorov&lt;br /&gt;
 |title=Unbiased estimates&lt;br /&gt;
 |journal=Izvestiya Akad. Nauk SSSR. Ser. Mat.&lt;br /&gt;
 |year=1950 |volume=14 |issue= |pages=303–326&lt;br /&gt;
 |mr=36479&lt;br /&gt;
}}&lt;br /&gt;
{{Refend}}&lt;br /&gt;
&lt;br /&gt;
==External links==&lt;br /&gt;
* {{SpringerEOM |title=Rao–Blackwell–Kolmogorov theorem |id=R/r077550 |first=M.S. |last=Nikulin}}&lt;br /&gt;
&lt;br /&gt;
{{DEFAULTSORT:Rao-Blackwell Theorem}}&lt;br /&gt;
[[Category:Statistical theorems]]&lt;br /&gt;
[[Category:Estimation theory]]&lt;/div&gt;</summary>
		<author><name>130.184.19.175</name></author>
	</entry>
	<entry>
		<id>https://en.formulasearchengine.com/w/index.php?title=Malmquist_bias&amp;diff=241221</id>
		<title>Malmquist bias</title>
		<link rel="alternate" type="text/html" href="https://en.formulasearchengine.com/w/index.php?title=Malmquist_bias&amp;diff=241221"/>
		<updated>2012-07-06T19:20:06Z</updated>

		<summary type="html">&lt;p&gt;130.184.60.127: The sentence stated &amp;quot;greater absolute magnitudes&amp;quot; when referring to brighter (i.e. smaller absolute magnitude) objects. I changed it accordingly.&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;My name is Denny and I am studying Arts and Sciences and Integrated International Studies at Croix / France.&amp;lt;br&amp;gt;&amp;lt;br&amp;gt;Have a look at my blog :: [http://www.ttumc.net/member/249229/ FIFA 15 coin hack]&lt;/div&gt;</summary>
		<author><name>130.184.60.127</name></author>
	</entry>
</feed>