# Uniformly most powerful test

Uniformly most powerful test

In statistical hypothesis testing, a uniformly most powerful (UMP) test is a hypothesis test which has the greatest power among all possible tests of a given size "α". For example, according to the Neyman-Pearson lemma, the likelihood-ratio test is UMP for testing simple (point) hypotheses.

Setting

Let $X$ denote a random vector (corresponding to the measurements), taken from a parametrized family of probability density functions or probability mass functions $f_\left\{ heta\right\}\left(x\right)$, which depends on the unknown deterministic parameter $heta in Theta$. The parameter space $Theta$ is partitioned into two disjoint sets $Theta_0$ and $Theta_1$. Let $H_0$ denote the hypothesis that $heta in Theta_0$, and let $H_1$ denote the hypothesis that $heta in Theta_1$.The binary test of hypotheses is performed using a test function $phi\left(x\right)$. : meaning that $H_1$ is in force if the measurement $X in R$ and that $H_0$ is in force if the measurement $X in A$. $A cup R$ is a disjoint covering of the measurement space.

Formal definition

A test function $phi\left(x\right)$ is UMP of size $alpha$ if for any other test function $phi\text{'}\left(x\right)$ we have::$sup_\left\{ hetainTheta_0\right\}; E_ hetaphi\text{'}\left(X\right)=alpha\text{'}leqalpha=sup_\left\{ hetainTheta_0\right\}; E_ hetaphi\left(X\right),$:

The Karlin-Rubin theorem

The Karlin-Rubin theorem can be regarded as an extension of the Neyman-Pearson lemma for composite hypotheses. Consider a scalar measurement having a probability density function parameterized by a scalar parameter "&theta;", and define the likelihood ratio $l\left(x\right) = f_\left\{ heta_1\right\}\left(x\right) / f_\left\{ heta_0\right\}\left(x\right)$.If $l\left(x\right)$ is monotone non-decreasing for any pair $heta_1 geq heta_0$ (meaning that the greater $x$ is, the more likely $H_1$ is), then the threshold test:::$E_\left\{ heta_0\right\}phi\left(X\right)=alpha$is the UMP test of size "α" for testing $H_0: heta leq heta_0 ext\left\{ vs. \right\} H_1: heta > heta_0$

Note that exactly the same test is also UMP for testing $H_0: heta = heta_0 ext\left\{ vs. \right\} H_1: heta > heta_0$

Important case: The exponential family

Although the Karlin-Rubin may seem weak because of its restriction to scalar parameter and scalar measurement, it turns out that there exist a host of problems for which the theorem holds. In particular, the one-dimensional exponential family of probability density functions or probability mass functions with $f_ heta\left(x\right) = c\left( heta\right)h\left(x\right)exp\left(pi\left( heta\right)T\left(x\right)\right)$ has a monotone non-decreasing likelihood ratio in the sufficient statistic "T"("x"), provided that $pi\left( heta\right)$ is non-decreasing.

Example

Let $X=\left(X_0 , X_1 ,dots , X_\left\{M-1\right\}\right)$ denote i.i.d. normally distributed $N$-dimensional random vectors with mean $heta m$ and covariance matrix $R$. We then have :$f_ heta \left(X\right) = \left(2 pi\right)^\left\{-M N / 2\right\} |R|^\left\{-M / 2\right\} exp left\left\{-frac\left\{1\right\}\left\{2\right\} sum_\left\{n=0\right\}^\left\{M-1\right\}\left(X_n - heta m\right)^T R^\left\{-1\right\}\left(X_n - heta m\right) ight\right\} =$:$= \left(2 pi\right)^\left\{-M N / 2\right\} |R|^\left\{-M / 2\right\} exp left\left\{-frac\left\{1\right\}\left\{2\right\} sum_\left\{n=0\right\}^\left\{M-1\right\}\left( heta^2 m^T R^\left\{-1\right\} m\right) ight\right\} cdot exp left\left\{-frac\left\{1\right\}\left\{2\right\} sum_\left\{n=0\right\}^\left\{M-1\right\}X_n^T R^\left\{-1\right\} X_n ight\right\} cdot exp left\left\{ heta m^T R^\left\{-1\right\} sum_\left\{n=0\right\}^\left\{M-1\right\}X_n ight\right\}$which is exactly in the form of the exponential family shown in the previous section, with the sufficient statistic being

: $T\left(X\right) = m^T R^\left\{-1\right\} sum_\left\{n=0\right\}^\left\{M-1\right\}X_n.$

Thus, we conclude that the test::$E_\left\{ heta_0\right\} phi \left(T\right) = alpha$

is the UMP test of size $alpha$ for testing $H_0: heta leq heta_0$ vs. $H_1: heta > heta_0$

Further discussion

Finally, we note that in general, UMP tests do not exist for vector parameters or for two-sided tests (a test in which one hypothesis lies on both sides of the alternative). Why is it so?

The reason is that in these situations, the most powerful test of a given size for one possible value of the parameter (e.g. for $heta_1$ where $heta_1 > heta_0$) is different than the most powerful test of the same size for a different value of the parameter (e.g. for $heta_2$ where $heta_2 < heta_0$). As a result, no test is Uniformly most powerful.

References

* L. L. Scharf, "Statistical Signal Processing", Addison-Wesley, 1991, section 4.7.

Wikimedia Foundation. 2010.

### Look at other dictionaries:

• Score test — A score test is a statistical test of a simple null hypothesis that a parameter of interest heta isequal to some particular value heta 0. It is the most powerful test when the true value of heta is close to heta 0.ingle parameter testThe… …   Wikipedia

• UMP — *Uniformly most powerful test, in statistical hypothesis testing *Union for a Popular Movement or UMP, the main French centre right political party *United Midwestern Promoters, a racing sanctioning body *Ulyanovsk Mechanical Plant *Heckler Koch… …   Wikipedia

• Statistical hypothesis testing — This article is about frequentist hypothesis testing which is taught in introductory statistics. For Bayesian hypothesis testing, see Bayesian inference. A statistical hypothesis test is a method of making decisions using data, whether from a… …   Wikipedia

• Exponential family — Not to be confused with the exponential distribution. Natural parameter links here. For the usage of this term in differential geometry, see differential geometry of curves. In probability and statistics, an exponential family is an important… …   Wikipedia

• List of statistics topics — Please add any Wikipedia articles related to statistics that are not already on this list.The Related changes link in the margin of this page (below search) leads to a list of the most recent changes to the articles listed below. To see the most… …   Wikipedia

• List of mathematics articles (U) — NOTOC U U duality U quadratic distribution U statistic UCT Mathematics Competition Ugly duckling theorem Ulam numbers Ulam spiral Ultraconnected space Ultrafilter Ultrafinitism Ultrahyperbolic wave equation Ultralimit Ultrametric space… …   Wikipedia

• Unsolved problems in statistics — There are many longstanding unsolved problems in mathematics for which a solution has still not yet been found. The unsolved problems in statistics are generally of a different flavor; according to John Tukey, difficulties in identifying problems …   Wikipedia

• Modern Language Aptitude Test — The Modern Language Aptitude Test was designed to predict a student’s likelihood of success and ease in learning a foreign language. The Modern Language Aptitude Test (MLAT) was developed to measure foreign language learning aptitude. Language… …   Wikipedia

• UMP (disambiguation) — UMP is a three letter acronym that could mean:* Union pour un Mouvement Populaire, (English: Union for a Popular Movement ), French ruling party. * Universal Machine Pistol, a submachine gun made by Heckler Koch * Upper Merion and Plymouth… …   Wikipedia

• Monotone likelihood ratio property — is a property of a family of probability distributions described by their probability density functions (PDFs). A family of density functions { f heta (x)} { hetain Theta} indexed by a parameter heta taking values in a set Theta is said to have… …   Wikipedia