|
|
| Line 1: |
Line 1: |
| In [[statistics]], the '''mean squared error (MSE)''' of an [[estimator]] is one of many ways to quantify the difference between values implied by an estimator and the true values of the quantity being estimated. MSE is a [[risk function]], corresponding to the [[expected value]] of the squared error loss or quadratic loss. MSE measures the [[expected value|average]] of the squares of the "errors." The error is the amount by which the value implied by the estimator differs from the quantity to be estimated. The difference occurs because of [[randomness]] or because the estimator [[Omitted-variable bias|doesn't account for information]] that could produce a more accurate estimate.<ref name="pointEstimation">{{cite book
| | Hello! My name is Jacqueline. <br>It is a little about myself: I live in Italy, my city of Forcella. <br>It's called often Eastern or cultural capital of TE. I've married 1 years ago.<br>I have two children - a son (Lashonda) and the daughter (Annette). We all like Gongoozling.<br><br>Feel free to surf to my site; [http://www.stewartsurfboards.com/wp-louboutin.php christian louboutin knockoffs] |
| |first1=E. L. |last1=Lehmann
| |
| |first2=George |last2=Casella
| |
| |title=Theory of Point Estimation
| |
| |publisher=Springer |location=New York
| |
| |year=1998 |edition=2nd
| |
| |isbn=0-387-98502-6 |mr=1639875
| |
| }}</ref>
| |
| | |
| The MSE is the second [[moment (mathematics)|moment]] (about the origin) of the error, and thus incorporates both the [[variance]] of the estimator and its [[Bias of an estimator|bias]]. For an [[unbiased estimator]], the MSE is the variance of the estimator. Like the variance, MSE has the same units of measurement as the square of the quantity being estimated. In an analogy to [[standard deviation]], taking the square root of MSE yields the root-mean-square error or [[root-mean-square deviation]] (RMSE or RMSD), which has the same units as the quantity being estimated; for an unbiased estimator, the RMSE is the square root of the variance, known as the [[standard deviation]].
| |
| | |
| ==Definition and basic properties==
| |
| | |
| If <math>\hat{Y}</math> is a vector of n predictions, and <math>Y</math> is the vector of the true values, then the (estimated) MSE of the predictor is:
| |
| <math>\operatorname{MSE}=\frac{1}{n}\sum_{i=1}^n(\hat{Y_i} - Y_i)^2.</math>
| |
| | |
| This is a known, computed quantity given a particular sample (and hence is sample-dependent).
| |
| | |
| The MSE of an [[estimator]] <math>\hat{\theta}</math> with respect to the unknown parameter <math>\theta</math> is defined as
| |
| | |
| :<math>\operatorname{MSE}(\hat{\theta})=\operatorname{E}\big[(\hat{\theta}-\theta)^2\big].</math>
| |
| | |
| This definition depends on the unknown parameter, and the MSE in this sense is a property of an estimator (of a method of obtaining an estimate).
| |
| | |
| The MSE is equal to the sum of the [[variance]] and the squared [[bias of an estimator|bias]] of the estimator or of the predictions. In the case of the MSE of an estimator,<ref name="wackerly">{{cite book
| |
| |first1=Dennis |last1=Wackerly
| |
| |first2=William|last2=Mendenhall
| |
| |first3=Richard L.|last2=Scheaffer
| |
| |title=Mathematical Statistics with Applications
| |
| |publisher=Thomson Higher Education|location=Belmont, CA, USA
| |
| |year=2008 |edition=7
| |
| |isbn=0-495-38508-5
| |
| }}</ref>
| |
| | |
| :<math>\operatorname{MSE}(\hat{\theta})=\operatorname{Var}(\hat{\theta})+ \left(\operatorname{Bias}(\hat{\theta},\theta)\right)^2.</math>
| |
| {{Collapse top|title=Proof}}
| |
| <math>\begin{align}\operatorname{MSE}(\hat{\theta})\equiv \mathbb{E}((\hat{\theta}-\theta)^2)&=
| |
| \mathbb{E}\left[\left(\hat{\theta}-\mathbb{E}(\hat\theta)+\mathbb{E}(\hat\theta)-\theta\right)^2\right]
| |
| \\ & =
| |
| \mathbb{E}\left[\left(\hat{\theta}-\mathbb{E}(\hat\theta)\right)^2 +2\left((\hat{\theta}-\mathbb{E}(\hat\theta))(\mathbb{E}(\hat\theta)-\theta)\right)+\left( \mathbb{E}(\hat\theta)-\theta \right)^2\right]
| |
| \\ & = \mathbb{E}\left[\left(\hat{\theta}-\mathbb{E}(\hat\theta)\right)^2\right]+2\mathbb{E}\left[(\hat{\theta}-\mathbb{E}(\hat\theta))(\mathbb{E}(\hat\theta)-\theta)\right]+\mathbb{E}\left[\left(\mathbb{E}(\hat\theta)-\theta\right)^2\right]
| |
| \\ & = \mathbb{E}\left[\left(\hat{\theta}-\mathbb{E}(\hat\theta)\right)^2\right]+2(\mathbb{E}(\hat\theta)-\theta)\overbrace{\mathbb{E}(\hat{\theta}-\mathbb{E}(\hat\theta))}^{=\mathbb{E}(\hat\theta)-\mathbb{E}(\hat\theta)=0}+\mathbb{E}\left[\left(\mathbb{E}(\hat\theta)-\theta\right)^2\right]
| |
| \\ & = \mathbb{E}\left[\left(\hat{\theta}-\mathbb{E}(\hat\theta)\right)^2\right]+\mathbb{E}\left[\left(\mathbb{E}(\hat\theta)-\theta\right)^2\right]
| |
| \\ & = \operatorname{Var}(\hat\theta)+ \operatorname{Bias}(\hat\theta,\theta)^2
| |
| \end{align}</math>
| |
| {{Collapse bottom}}
| |
| | |
| The MSE thus assesses the quality of an estimator or set of predictions in terms of its variation and degree of bias.
| |
| | |
| Since MSE is an expectation, it is not a random variable. It may be a function of the unknown parameter <math>\theta</math>, but it does not depend on any random quantities. However, when MSE is computed for a particular estimator of <math>\theta</math> the true value of which is not known, it will be subject to estimation error. Thus, any estimation of the MSE on the basis of an estimated parameter is in fact a random variable. {{cn|date=February 2012}}
| |
| | |
| ==Regression==
| |
| | |
| In regression analysis, the term ''mean squared error'' is sometimes used to refer to the unbiased estimate of error variance: the [[residual sum of squares]] divided by the number of [[Degrees of freedom (statistics)|degrees of freedom]]. This definition for a known, computed quantity differs from the above definition for the computed MSE of a predictor in that a different denominator is used. The denominator is the sample size reduced by the number of model parameters estimated from the same data, ''(n-p)'' for ''p'' [[regressor]]s or ''(n-p-1)'' if an intercept is used.<ref>Steel, R.G.D, and Torrie, J. H., ''Principles and Procedures of Statistics with Special Reference to the Biological Sciences.'', [[McGraw Hill]], 1960, page 288.</ref> For more details, see [[errors and residuals in statistics]]. Note that, although the MSE is not an unbiased estimator of the error variance, it is consistent, given the consistency of the predictor.
| |
| | |
| Also in regression analysis, "mean squared error", often referred to as [[mean squared prediction error]] or "out-of-sample mean squared error", can refer to the mean value of the squared deviations of the predictions from the true values, over an out-of-sample test space, generated by a model estimated over a particular sample space. This also is a known, computed quantity, and it varies by sample and by out-of-sample test space...
| |
| | |
| ==Examples==
| |
| | |
| Suppose we have a random sample of size ''n'' from a population, <math>X_1,\dots,X_n</math>. Suppose the sample units were chosen with replacement. That is, the ''n'' units are selected one at a time, and previously selected units are still eligible for selection for all ''n'' draws. The usual estimator for the mean is the sample average
| |
| | |
| :<math>\overline{X}=\frac{1}{n}\sum_{i=1}^n X_i </math>
| |
| | |
| which has an expected value equal to the true mean μ (so it is unbiased) and a mean square error of
| |
| | |
| :<math>\operatorname{MSE}(\overline{X})=\operatorname{E}((\overline{X}-\mu)^2)=\left(\frac{\sigma}{\sqrt{n}}\right)^2= \frac{\sigma^2}{n}</math>
| |
| | |
| where <math>\sigma^2</math> is the population variance.
| |
| | |
| | |
| For a [[Gaussian distribution]] this is the [[best unbiased estimator]] (that is, it has the lowest MSE among all unbiased estimators), but not, say, for a [[Uniform distribution (continuous)|uniform distribution]].
| |
| | |
| The usual estimator for the variance is
| |
| | |
| :<math>S^2_{n-1} = \frac{1}{n-1}\sum_{i=1}^n\left(X_i-\overline{X}\,\right)^2
| |
| =\frac{1}{n-1}\left(\sum_{i=1}^n X_i^2-n\overline{X}^2\right).</math>
| |
| | |
| This is unbiased (its expected value is <math>\sigma^2</math>), and its MSE is<ref>{{cite book
| |
| | last = Mood |first=A.
| |
| | last2 = Graybill |first2=F.
| |
| |last3=Boes |first3=D.
| |
| | title = Introduction to the Theory of Statistics
| |
| |page=229
| |
| | edition = 3rd
| |
| | publisher = McGraw-Hill
| |
| | year = 1974
| |
| }}</ref>
| |
| | |
| :<math>\begin{align}\operatorname{MSE}(S^2_{n-1})&= \frac{1}{n} \left(\mu_4-\frac{n-3}{n-1}\sigma^4\right) \\
| |
| &=\frac{1}{n} \left(\gamma_2+\frac{2n}{n-1}\right)\sigma^4,\end{align}</math>
| |
| | |
| where <math>\mu_4</math> is the fourth [[central moment]] of the distribution or population and <math>\gamma_2=\mu_4/\sigma^4-3</math> is the [[excess kurtosis]].
| |
| | |
| However, one can use other estimators for <math>\sigma^2</math> which are proportional to <math>S^2_{n-1}</math>, and an appropriate choice can always give a lower mean square error. If we define
| |
| | |
| :<math>\begin{align}S^2_a &= \frac{n-1}{a}S^2_{n-1}\\
| |
| &= \frac{1}{a}\sum_{i=1}^n\left(X_i-\overline{X}\,\right)^2\end{align}</math>
| |
| | |
| then the MSE is
| |
| | |
| :<math>\begin{align}
| |
| \operatorname{MSE}(S^2_a)&=\operatorname{E}\left(\left(\frac{n-1}{a} S^2_{n-1}-\sigma^2\right)^2 \right) \\
| |
| &=\frac{n-1}{n a^2}[(n-1)\gamma_2+n^2+n]\sigma^4-\frac{2(n-1)}{a}\sigma^4+\sigma^4
| |
| \end{align}</math>
| |
| | |
| This is minimized when
| |
| | |
| :<math>a=\frac{(n-1)\gamma_2+n^2+n}{n} = n+1+\frac{n-1}{n}\gamma_2.</math>
| |
| | |
| For a [[Gaussian distribution]], where <math>\gamma_2=0</math>, this means the MSE is minimized when dividing the sum by <math>a=n+1</math>, whereas for a [[Bernoulli distribution]] with ''p'' = 1/2 (a coin flip), <math>\gamma_2=-2</math>, the MSE is minimized for <math>a=n-1+2/n</math>. (Note that this particular case of the Bernoulli distribution has the lowest possible [[excess kurtosis]]; this can be proved by [[Jensen's inequality]] as follows. The fourth [[central moment]] is an upper bound for the square of variance, so that the least value for their ratio is one, therefore, the least value for the [[excess kurtosis]] is -2, achieved, for instance, by a Bernoulli with p=1/2.) So no matter what the kurtosis, we get a "better" estimate (in the sense of having a lower MSE) by scaling down the unbiased estimator a little bit. Even among unbiased estimators, if the distribution is not Gaussian the best (minimum mean square error) estimator of the variance may not be <math>S^2_{n-1}.</math>
| |
| | |
| The following table gives several estimators of the true parameters of the population, μ and σ<sup>2</sup>, for the Gaussian case.<ref>{{cite book
| |
| | last = DeGroot
| |
| | first = Morris H.
| |
| | authorlink = Morris H. DeGroot
| |
| | title = Probability and Statistics
| |
| | edition = 2nd
| |
| | publisher = Addison-Wesley
| |
| | year = 1980
| |
| | ref = degroot }}</ref>
| |
| | |
| {| class="wikitable"
| |
| ! True value !! Estimator !! Mean squared error
| |
| |-
| |
| | θ = μ || <math>\hat{\theta}</math> = the unbiased estimator of the [[population mean]], <math>\overline{X}=\frac{1}{n}\sum_{i=1}^n(X_i)</math> || <math>\operatorname{MSE}(\overline{X})=\operatorname{E}((\overline{X}-\mu)^2)=\left(\frac{\sigma}{\sqrt{n}}\right)^2</math>
| |
| |-
| |
| | θ = σ<sup>2</sup> || <math>\hat{\theta}</math> = the unbiased estimator of the [[population variance]], <math>S^2_{n-1} = \frac{1}{n-1}\sum_{i=1}^n\left(X_i-\overline{X}\,\right)^2</math> || <math>\operatorname{MSE}(S^2_{n-1})=\operatorname{E}((S^2_{n-1}-\sigma^2)^2)=\frac{2}{n - 1}\sigma^4</math>
| |
| |-
| |
| | θ = σ<sup>2</sup> || <math>\hat{\theta}</math> = the biased estimator of the [[population variance]], <math>S^2_{n} = \frac{1}{n}\sum_{i=1}^n\left(X_i-\overline{X}\,\right)^2</math> || <math>\operatorname{MSE}(S^2_{n})=\operatorname{E}((S^2_{n}-\sigma^2)^2)=\frac{2n - 1}{n^2}\sigma^4</math>
| |
| |-
| |
| | θ = σ<sup>2</sup> || <math>\hat{\theta}</math> = the biased estimator of the [[population variance]], <math>S^2_{n+1} = \frac{1}{n+1}\sum_{i=1}^n\left(X_i-\overline{X}\,\right)^2</math> || <math>\operatorname{MSE}(S^2_{n+1})=\operatorname{E}((S^2_{n+1}-\sigma^2)^2)=\frac{2}{n + 1}\sigma^4</math>
| |
| |}
| |
| | |
| Note that:
| |
| #The MSEs shown for the variance estimators assume <math>X_i \sim \operatorname{N}(\mu,\sigma^2)</math> [[Independent and identically-distributed random variables|i.i.d.]] so that <math>\frac{(n-1)S^2_{n-1}}{\sigma^2}\sim \chi^2_{n-1}</math>. The result for <math>S^2_{n-1}</math> follows easily from the <math>\chi^2_{n-1}</math> variance that is <math>2n-2</math>.
| |
| #Unbiased estimators may not produce estimates with the smallest total variation (as measured by MSE): the MSE of <math>S^2_{n-1}</math> is larger than that of <math>S^2_{n+1}</math> or <math>S^2_n</math>.
| |
| #Estimators with the smallest total variation may produce biased estimates: <math>S^2_{n+1}</math> typically underestimates σ<sup>2</sup> by <math>\frac{2}{n}\sigma^2</math>
| |
| | |
| ==Interpretation==
| |
| | |
| An MSE of zero, meaning that the estimator <math>\hat{\theta}</math> predicts observations of the parameter <math>\theta</math> with perfect accuracy, is the ideal, but is practically never possible.
| |
| | |
| Values of MSE may be used for comparative purposes. Two or more [[statistical model]]s may be compared using their MSEs as a measure of how well they explain a given set of observations: An unbiased estimator (estimated from a statistical model) with the smallest variance among all unbiased estimators is the best prediction in the sense that it minimises the variance and is called the [[best unbiased estimator]] or MVUE (Minimum Variance Unbiased Estimator).
| |
| | |
| Both [[linear regression]] techniques such as [[analysis of variance]] estimate the MSE as part of the analysis and use the estimated MSE to determine the [[statistical significance]] of the factors or predictors under study. The goal of [[experimental design]] is to construct experiments in such a way that when the observations are analyzed, the MSE is close to zero relative to the magnitude of at least one of the estimated treatment effects.
| |
| | |
| MSE is also used in several [[stepwise regression]] techniques as part of the determination as to how many predictors from a candidate set to include in a model for a given set of observations.
| |
| | |
| ==Applications==
| |
| | |
| *Minimizing MSE is a key criterion in selecting estimators: see [[minimum mean-square error]]. Among unbiased estimators, minimizing the MSE is equivalent to minimizing the variance, and the estimator that does this is the [[minimum variance unbiased estimator]]. However, a biased estimator may have lower MSE; see [[estimator bias]].
| |
| | |
| *In [[statistical modelling]] the MSE, representing the difference between the actual observations and the observation values predicted by the model, is used to determine the extent to which the model fits the data and whether the removal or some explanatory variables, simplifying the model, is possible without significantly harming the model's predictive ability.
| |
| | |
| ==As a loss function==
| |
| | |
| Squared error loss is one of the most widely used [[loss function]]s in statistics, though its widespread use stems more from mathematical convenience than considerations of actual loss in applications. [[Carl Friedrich Gauss]], who introduced the use of mean squared error, was aware of its arbitrariness and was in agreement with objections to it on these grounds.<ref name="pointEstimation" /> The mathematical benefits of mean squared error are particularly evident in its use at analyzing the performance of [[linear regression]], as it allows one to partition the variation in a dataset into variation explained by the model and variation explained by randomness.
| |
| | |
| ;Criticism | |
| The use of mean squared error without question has been criticized by the [[decision theory|decision theorist]] [[James Berger (statistician)|James Berger]]. Mean squared error is the negative of the expected value of one specific [[utility function]], the quadratic utility function, which may not be the appropriate utility function to use under a given set of circumstances. There are, however, some scenarios where mean squared error can serve as a good approximation to a loss function occurring naturally in an application.<ref>{{cite book
| |
| |title=Statistical decision theory and Bayesian Analysis
| |
| |first=James O. |last=Berger |authorlink=James Berger (statistician)
| |
| |year=1985
| |
| |edition=2nd
| |
| |publisher=Springer-Verlag |location=New York
| |
| |ISBN=0-387-96098-8 |mr=0804611
| |
| |chapter=2.4.2 Certain Standard Loss Functions |page=60
| |
| }}</ref>
| |
| | |
| Like [[variance]], mean squared error has the disadvantage of heavily weighting [[outliers]].<ref>Sergio Bermejo, Joan Cabestany (2001) "[http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6T08-43PS3GC-1&_user=483692&_coverDate=12%2F31%2F2001&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000022720&_version=1&_urlVersion=0&_userid=483692&md5=8586e409694e1b50da3aa3c6fce18cb8 Oriented principal component analysis for large margin classifiers]", ''Neural Networks'', 14 (10), 1447–1461.</ref> This is a result of the squaring of each term, which effectively weights large errors more heavily than small ones. This property, undesirable in many applications, has led researchers to use alternatives such as the [[mean absolute error]], or those based on the [[median]].
| |
| | |
| ==See also==
| |
| *[[James–Stein estimator]]
| |
| *[[Hodges' estimator]]
| |
| *[[Mean percentage error]]
| |
| *[[Mean square weighted deviation]]
| |
| *[[Mean squared displacement]]
| |
| *[[Mean squared prediction error]]
| |
| *[[Minimum mean squared error|Minimum mean squared error estimator]]
| |
| *[[Peak signal-to-noise ratio]]
| |
| *[[Root mean square deviation]]
| |
| *[[Squared deviations]]
| |
| | |
| ==Notes==
| |
| {{reflist}}
| |
| | |
| [[Category:Estimation theory]]
| |
| [[Category:Point estimation performance]]
| |
| [[Category:Statistical deviation and dispersion]]
| |
| [[Category:Loss functions]]
| |
| [[Category:Least squares]]
| |