The median of a set of numbers is the value separating the higher half from the lower half of a data sample, a population, or a probability distribution. For a data set, it may be thought of as the “middle" value. The basic feature of the median in describing data compared to the mean (often simply described as the "average") is that it is not skewed by a small proportion of extremely large or small values, and therefore provides a better representation of the center. Median income, for example, may be a better way to describe the center of the income distribution because increases in the largest incomes alone have no effect on the median. For this reason, the median is of central importance in robust statistics.
The median of a finite list of numbers is the "middle" number, when those numbers are listed in order from smallest to greatest.
If the data set has an odd number of observations, the middle one is selected (after arranging in ascending order). For example, the following list of seven numbers,
1, 3, 3, 6, 7, 8, 9has the median of 6, which is the fourth value.
If the data set has an even number of observations, there is no distinct middle value and the median is usually defined to be the arithmetic mean of the two middle values.[1] For example, this data set of 8 numbers
1, 2, 3, 4, 5, 6, 8, 9has a median value of 4.5, that is
(4+5)/2
In general, with this convention, the median can be defined as follows: For a data set
x
n
if
n
\operatorname{med}(x)=x(n
if
n
\operatorname{med}(x)=
| x(n/2)+x((n/2)+1) | |
| 2 |
| Midrange | Midway point between the minimum and the maximum of a data set | 1, 2, 2, 3, 4, 7, 9 | 5 | |
| Arithmetic mean | Sum of values of a data set divided by number of values: | 4 | ||
| Median | Middle value separating the greater and lesser halves of a data set | 1, 2, 2, 3, 4, 7, 9 | 3 | |
| Mode | Most frequent value in a data set | 1, 2, 2, 3, 4, 7, 9 | 2 |
Formally, a median of a population is any value such that at least half of the population is less than or equal to the proposed median and at least half is greater than or equal to the proposed median. As seen above, medians may not be unique. If each set contains more than half the population, then some of the population is exactly equal to the unique median.
The median is well-defined for any ordered (one-dimensional) data and is independent of any distance metric. The median can thus be applied to school classes which are ranked but not numerical (e.g. working out a median grade when student test scores are graded from F to A), although the result might be halfway between classes if there is an even number of classes. (For odd number classes, one specific class is determined as the median.)
A geometric median, on the other hand, is defined in any number of dimensions. A related concept, in which the outcome is forced to correspond to a member of the sample, is the medoid.
There is no widely accepted standard notation for the median, but some authors represent the median of a variable x as med(x), x͂,[2] as μ1/2, or as M.[2] [3] In any of these cases, the use of these or other symbols for the median needs to be explicitly defined when they are introduced.
The median is a special case of other ways of summarizing the typical values associated with a statistical distribution: it is the 2nd quartile, 5th decile, and 50th percentile.
The median can be used as a measure of location when one attaches reduced importance to extreme values, typically because a distribution is skewed, extreme values are not known, or outliers are untrustworthy, i.e., may be measurement or transcription errors.
For example, consider the multiset
1, 2, 2, 2, 3, 14.
The median is 2 in this case, as is the mode, and it might be seen as a better indication of the center than the arithmetic mean of 4, which is larger than all but one of the values. However, the widely cited empirical relationship that the mean is shifted "further into the tail" of a distribution than the median is not generally true. At most, one can say that the two statistics cannot be "too far" apart; see below.[4]
As a median is based on the middle data in a set, it is not necessary to know the value of extreme results in order to calculate it. For example, in a psychology test investigating the time needed to solve a problem, if a small number of people failed to solve the problem at all in the given time a median can still be calculated.[5]
Because the median is simple to understand and easy to calculate, while also a robust approximation to the mean, the median is a popular summary statistic in descriptive statistics. In this context, there are several choices for a measure of variability: the range, the interquartile range, the mean absolute deviation, and the median absolute deviation.
For practical purposes, different measures of location and dispersion are often compared on the basis of how well the corresponding population values can be estimated from a sample of data. The median, estimated using the sample median, has good properties in this regard. While it is not usually optimal if a given population distribution is assumed, its properties are always reasonably good. For example, a comparison of the efficiency of candidate estimators shows that the sample mean is more statistically efficient when—and only when— data is uncontaminated by data from heavy-tailed distributions or from mixtures of distributions. Even then, the median has a 64% efficiency compared to the minimum-variance mean (for large normal samples), which is to say the variance of the median will be ~50% greater than the variance of the mean.[6] [7]
For any real-valued probability distribution with cumulative distribution function F, a median is defined as any real number m that satisfies the inequalities(cf. the drawing in the definition of expected value for arbitrary real-valued random variables). An equivalent phrasing uses a random variable X distributed according to F:
Note that this definition does not require X to have an absolutely continuous distribution (which has a probability density function f), nor does it require a discrete one. In the former case, the inequalities can be upgraded to equality: a median satisfiesand
Any probability distribution on the real number set
\R
The medians of certain types of distributions can be easily calculated from their parameters; furthermore, they exist even for some distributions lacking a well-defined mean, such as the Cauchy distribution:
The mean absolute error of a real variable c with respect to the random variable X is
E(\left|X-c\right|)
More generally, a median is defined as a minimum of
E(|X-c|-|X|),
This optimization-based definition of the median is useful in statistical data-analysis, for example, in k-medians clustering.
If the distribution has finite variance, then the distance between the median
\tilde{X}
\bar{X}
This bound was proved by Book and Sher in 1979 for discrete samples,[11] and more generally by Page and Murty in 1982.[12] In a comment on a subsequent proof by O'Cinneide,[13] Mallows in 1991 presented a compact proof that uses Jensen's inequality twice,[14] as follows. Using |·| for the absolute value, we have
\begin{align} |\mu-m|=|\operatorname{E}(X-m)|&\leq\operatorname{E}(|X-m|)\\ &\leq\operatorname{E}(|X-\mu|)\\ &\leq\sqrt{\operatorname{E}\left((X-\mu)2\right)}=\sigma. \end{align}
The first and third inequalities come from Jensen's inequality applied to the absolute-value function and the square function, which are each convex. The second inequality comes from the fact that a median minimizes the absolute deviation function
a\mapsto\operatorname{E}(|X-a|)
Mallows's proof can be generalized to obtain a multivariate version of the inequality[15] simply by replacing the absolute value with a norm:
\|\mu-m\|\leq\sqrt{\operatorname{E}\left(\|X-\mu\|2\right)}=\sqrt{\operatorname{trace}\left(\operatorname{var}(X)\right)}
where m is a spatial median, that is, a minimizer of the function
a\mapsto\operatorname{E}(\|X-a\|).
An alternative proof uses the one-sided Chebyshev inequality; it appears in an inequality on location and scale parameters. This formula also follows directly from Cantelli's inequality.[18]
For the case of unimodal distributions, one can achieve a sharper bound on the distance between the median and the mean:
\left|\tilde{X}-\bar{X}\right|\le\left(
| 3 | |
| 5 |
| ||||
| \right) |
\sigma ≈ 0.7746\sigma
A similar relation holds between the median and the mode:
\left|\tilde{X}-mode\right|\le
| ||||
| 3 |
\sigma ≈ 1.732\sigma.
A typical heuristic is that positively skewed distributions have mean > median. This is true for all members of the Pearson distribution family. However this is not always true. For example, the Weibull distribution family has members with positive mean, but mean < median. Violations of the rule are particularly common for discrete distributions. For example, any Poisson distribution has positive skew, but its mean < median whenever
\mu\bmod1>ln2
When the distribution has a monotonically decreasing probability density, then the median is less than the mean, as shown in the figure.
Jensen's inequality states that for any random variable X with a finite expectation E[''X''] and for any convex function f
f[E(x)]\leE[f(x)]
This inequality generalizes to the median as well. We say a function is a C function if, for any t,
f-1\left((-infty,t]\right)=\{x\inR\midf(x)\let\}
f(\operatorname{med}[X])\le\operatorname{med}[f(X)]
If the medians are not unique, the statement holds for the corresponding suprema.[22]
Even though comparison-sorting n items requires operations, selection algorithms can compute the th-smallest of items with only operations. This includes the median, which is the th order statistic (or for an even number of samples, the arithmetic mean of the two middle order statistics).[23]
Selection algorithms still have the downside of requiring memory, that is, they need to have the full sample (or a linear-sized portion of it) in memory. Because this, as well as the linear time requirement, can be prohibitive, several estimation procedures for the median have been developed. A simple one is the median of three rule, which estimates the median as the median of a three-element subsample; this is commonly used as a subroutine in the quicksort sorting algorithm, which uses an estimate of its input's median. A more robust estimator is Tukey's ninther, which is the median of three rule applied with limited recursion:[24] if is the sample laid out as an array, and
,
then
The remedian is an estimator for the median that requires linear time but sub-linear memory, operating in a single pass over the sample.[25]
The distributions of both the sample mean and the sample median were determined by Laplace.[26] The distribution of the sample median from a population with a density function
f(x)
\mu
| 1 | |
| 4nf(m)2 |
where
m
f(x)
n
Samplemedian\siml{N}\left(\mu=m,
| ||||
| \sigma |
\right)
A modern proof follows below. Laplace's result is now understood as a special case of the asymptotic distribution of arbitrary quantiles.
For normal samples, the density is
f(m)=1/\sqrt{2\pi\sigma2}
({\pi}/{2}) ⋅ (\sigma2/n).
We take the sample size to be an odd number
N=2n+1
F(v)
f(v)
1-F(v)
v
| \Pr[\operatorname{med}=v]dv= | (2n+1)! |
| n!n! |
F(v)n(1-F(v))nf(v)dv
Now we introduce the beta function. For integer arguments
\alpha
\beta
\Beta(\alpha,\beta)=
| (\alpha-1)!(\beta-1)! | |
| (\alpha+\beta-1)! |
f(v)dv=dF(v)
\alpha
\beta
n+1
| F(v)n(1-F(v))n | |
| \Beta(n+1,n+1) |
dF(v)
Hence the density function of the median is a symmetric beta distribution pushed forward by
F
1/(4(N+2))
| 1 | |
| 4(N+2)f(m)2 |
The additional 2 is negligible in the limit.
In practice, the functions
f
F
| v | 0 | 0.5 | 1 | 1.5 | 2 | 2.5 | 3 | 3.5 | 4 | 4.5 | 5 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| f(v) | 0.000 | 0.008 | 0.010 | 0.013 | 0.083 | 0.108 | 0.328 | 0.220 | 0.202 | 0.023 | 0.005 |
| F(v) | 0.000 | 0.008 | 0.018 | 0.031 | 0.114 | 0.222 | 0.550 | 0.770 | 0.972 | 0.995 | 1.000 |
Because the observations are discrete-valued, constructing the exact distribution of the median is not an immediate translation of the above expression for
\Pr(\operatorname{med}=v)
\Pr(\operatorname{med}=v)=
| n | |
| \sum | |
| i=0 |
| n | |
| \sum | |
| k=0 |
| N! | |
| i!(N-i-k)!k! |
F(v-1)i(1-F(v))kf(v)N-i-k
Here, i is the number of points strictly less than the median and k the number strictly greater.
Using these preliminaries, it is possible to investigate the effect of sample size on the standard errors of the mean and median. The observed mean is 3.16, the observed raw median is 3 and the observed interpolated median is 3.174. The following table gives some comparison statistics.
| 3 | 9 | 15 | 21 | ||
|---|---|---|---|---|---|
| Expected value of median | 3.198 | 3.191 | 3.174 | 3.161 | |
| Standard error of median (above formula) | 0.482 | 0.305 | 0.257 | 0.239 | |
| Standard error of median (asymptotic approximation) | 0.879 | 0.508 | 0.393 | 0.332 | |
| Standard error of mean | 0.421 | 0.243 | 0.188 | 0.159 |
The value of
(2f(x))-2
n-1/2(\nu-m)
\nu
k
| ||||
| n |
The efficiency of the sample median, measured as the ratio of the variance of the mean to the variance of the median, depends on the sample size and on the underlying population distribution. For a sample of size
N=2n+1
| 2 | |
| \pi |
| N+2 | |
| N |
The efficiency tends to
| 2 | |
| \pi |
N
In other words, the relative variance of the median will be
\pi/2 ≈ 1.57
| ||||
| (\pi/2) |
≈ 1.25
\sigma/\sqrt{n}
For univariate distributions that are symmetric about one median, the Hodges–Lehmann estimator is a robust and highly efficient estimator of the population median.[34]
If data is represented by a statistical model specifying a particular family of probability distributions, then estimates of the median can be obtained by fitting that family of probability distributions to the data and calculating the theoretical median of the fitted distribution. Pareto interpolation is an application of this when the population is assumed to have a Pareto distribution.
Previously, this article discussed the univariate median, when the sample or population had one-dimension. When the dimension is two or higher, there are multiple concepts that extend the definition of the univariate median; each such multivariate median agrees with the univariate median when the dimension is exactly one.[35] [36] [37]
The marginal median is defined for vectors defined with respect to a fixed set of coordinates. A marginal median is defined to be the vector whose components are univariate medians. The marginal median is easy to compute, and its properties were studied by Puri and Sen.[38]
The geometric median of a discrete set of sample points
x1,\ldotsxN
\hat\mu=\underset{\mu\inRm}{\operatorname{argmin}}
| N | |
| \sum | |
| n=1 |
\left\|\mu-xn\right\|2
In contrast to the marginal median, the geometric median is equivariant with respect to Euclidean similarity transformations such as translations and rotations.
If the marginal medians for all coordinate systems coincide, then their common location may be termed the "median in all directions".[39] This concept is relevant to voting theory on account of the median voter theorem. When it exists, the median in all directions coincides with the geometric median (at least for discrete distributions).
When dealing with a discrete variable, it is sometimes useful to regard the observed values as being midpoints of underlying continuous intervals. An example of this is a Likert scale, on which opinions or preferences are expressed on a scale with a set number of possible responses. If the scale consists of the positive integers, an observation of 3 might be regarded as representing the interval from 2.50 to 3.50. It is possible to estimate the median of the underlying variable. If, say, 22% of the observations are of value 2 or below and 55.0% are of 3 or below (so 33% have the value 3), then the median
m
x
F(x)
w
f(x)
mint=m+w\left[
| 1 | |
| 2 |
-
| ||||||
| f(m) |
\right].
Alternatively, if in an observed sample there are
k
j
i
mint=m+
| w | \left[ | |
| 2 |
| k-i | |
| j\right]. |
See main article: Pseudomedian. For univariate distributions that are symmetric about one median, the Hodges–Lehmann estimator is a robust and highly efficient estimator of the population median; for non-symmetric distributions, the Hodges–Lehmann estimator is a robust and highly efficient estimator of the population pseudo-median, which is the median of a symmetrized distribution and which is close to the population median.[40] The Hodges–Lehmann estimator has been generalized to multivariate distributions.[41]
The Theil–Sen estimator is a method for robust linear regression based on finding medians of slopes.[42]
The median filter is an important tool of image processing, that can effectively remove any salt and pepper noise from grayscale images.
See main article: k-medians clustering. In cluster analysis, the k-medians clustering algorithm provides a way of defining clusters, in which the criterion of maximising the distance between cluster-means that is used in k-means clustering, is replaced by maximising the distance between cluster-medians.
This is a method of robust regression. The idea dates back to Wald in 1940 who suggested dividing a set of bivariate data into two halves depending on the value of the independent parameter
x
y
x
Nair and Shrivastava in 1942 suggested a similar idea but instead advocated dividing the sample into three equal parts before calculating the means of the subsamples.[44] Brown and Mood in 1951 proposed the idea of using the medians of two subsamples rather the means.[45] Tukey combined these ideas and recommended dividing the sample into three equal size subsamples and estimating the line based on the medians of the subsamples.[46]
Any mean-unbiased estimator minimizes the risk (expected loss) with respect to the squared-error loss function, as observed by Gauss. A median-unbiased estimator minimizes the risk with respect to the absolute-deviation loss function, as observed by Laplace. Other loss functions are used in statistical theory, particularly in robust statistics.
The theory of median-unbiased estimators was revived by George W. Brown in 1947:
Further properties of median-unbiased estimators have been reported. Median-unbiased estimators are invariant under one-to-one transformations.
There are methods of constructing median-unbiased estimators that are optimal (in a sense analogous to the minimum-variance property for mean-unbiased estimators). Such constructions exist for probability distributions having monotone likelihood-functions.[47] [48] One such procedure is an analogue of the Rao–Blackwell procedure for mean-unbiased estimators: The procedure holds for a smaller class of probability distributions than does the Rao—Blackwell procedure but for a larger class of loss functions.[49]
Scientific researchers in the ancient near east appear not to have used summary statistics altogether, instead choosing values that offered maximal consistency with a broader theory that integrated a wide variety of phenomena.[50] Within the Mediterranean (and, later, European) scholarly community, statistics like the mean are fundamentally a medieval and early modern development. (The history of the median outside Europe and its predecessors remains relatively unstudied.)
The idea of the median appeared in the 6th century in the Talmud, in order to fairly analyze divergent appraisals.[51] [52] However, the concept did not spread to the broader scientific community.
Instead, the closest ancestor of the modern median is the mid-range, invented by Al-Biruni[53] [54] Transmission of his work to later scholars is unclear. He applied his technique to assaying currency metals, but, after he published his work, most assayers still adopted the most unfavorable value from their results, lest they appear to cheat. [55] However, increased navigation at sea during the Age of Discovery meant that ship's navigators increasingly had to attempt to determine latitude in unfavorable weather against hostile shores, leading to renewed interest in summary statistics. Whether rediscovered or independently invented, the mid-range is recommended to nautical navigators in Harriot's "Instructions for Raleigh's Voyage to Guiana, 1595".
The idea of the median may have first appeared in Edward Wright's 1599 book Certaine Errors in Navigation on a section about compass navigation.[56] Wright was reluctant to discard measured values, and may have felt that the median — incorporating a greater proportion of the dataset than the mid-range — was more likely to be correct. However, Wright did not give examples of his technique's use, making it hard to verify that he described the modern notion of median. The median (in the context of probability) certainly appeared in the correspondence of Christiaan Huygens, but as an example of a statistic that was inappropriate for actuarial practice.
The earliest recommendation of the median dates to 1757, when Roger Joseph Boscovich developed a regression method based on the L1 norm and therefore implicitly on the median.[57] In 1774, Laplace made this desire explicit: he suggested the median be used as the standard estimator of the value of a posterior PDF. The specific criterion was to minimize the expected magnitude of the error;
|\alpha-\alpha*|
\alpha*
\alpha
(\alpha-\alpha*)2
Antoine Augustin Cournot in 1843 was the first[60] to use the term median (valeur médiane) for the value that divides a probability distribution into two equal halves. Gustav Theodor Fechner used the median (Centralwerth) in sociological and psychological phenomena.[61] It had earlier been used only in astronomy and related fields. Gustav Fechner popularized the median into the formal analysis of data, although it had been used previously by Laplace, and the median appeared in a textbook by F. Y. Edgeworth.[62] Francis Galton used the English term median in 1881,[63] [64] having earlier used the terms middle-most value in 1869, and the medium in 1880.[65] [66]
Statisticians encouraged the use of medians intensely throughout the 19th century for its intuitive clarity. However, the notion of median does not lend itself to the theory of higher moments as well as the arithmetic mean does, and is much harder to compute. As a result, the median was steadily supplanted as a notion of generic average by the arithmetic mean during the 20th century.