In statistics and in probability theory, distance correlation or distance covariance is a measure of dependence between two paired random vectors of arbitrary, not necessarily equal, dimension. The population distance correlation coefficient is zero if and only if the random vectors are independent. Thus, distance correlation measures both linear and nonlinear association between two random variables or random vectors. This is in contrast to Pearson's correlation, which can only detect linear association between two random variables.

Distance correlation can be used to perform a statistical test of dependence with a permutation test. One first computes the distance correlation (involving the re-centering of Euclidean distance matrices) between two random vectors, and then compares this value to the distance correlations of many shuffles of the data.

Several sets of (xy) points, with the distance correlation coefficient of x and y for each set. Compare to the graph on correlation

Background

edit

The classical measure of dependence, the Pearson correlation coefficient,[1] is mainly sensitive to a linear relationship between two variables. Distance correlation was introduced in 2005 by Gábor J. Székely in several lectures to address this deficiency of Pearson's correlation, namely that it can easily be zero for dependent variables. Correlation = 0 (uncorrelatedness) does not imply independence while distance correlation = 0 does imply independence. The first results on distance correlation were published in 2007 and 2009.[2][3] It was proved that distance covariance is the same as the Brownian covariance.[3] These measures are examples of energy distances.

The distance correlation is derived from a number of other quantities that are used in its specification, specifically: distance variance, distance standard deviation, and distance covariance. These quantities take the same roles as the ordinary moments with corresponding names in the specification of the Pearson product-moment correlation coefficient.

Definitions

edit

Distance covariance

edit

Let us start with the definition of the sample distance covariance. Let (XkYk), k = 1, 2, ..., n be a statistical sample from a pair of real valued or vector valued random variables (XY). First, compute the n by n distance matrices (aj, k) and (bj, k) containing all pairwise distances

 

where ||⋅ ||denotes Euclidean norm. Then take all doubly centered distances

 

where   is the j-th row mean,   is the k-th column mean, and   is the grand mean of the distance matrix of the X sample. The notation is similar for the b values. (In the matrices of centered distances (Aj, k) and (Bj,k) all rows and all columns sum to zero.) The squared sample distance covariance (a scalar) is simply the arithmetic average of the products Aj, k Bj, k:

 

The statistic Tn = n dCov2n(X, Y) determines a consistent multivariate test of independence of random vectors in arbitrary dimensions. For an implementation see dcov.test function in the energy package for R.[4]

The population value of distance covariance can be defined along the same lines. Let X be a random variable that takes values in a p-dimensional Euclidean space with probability distribution μ and let Y be a random variable that takes values in a q-dimensional Euclidean space with probability distribution ν, and suppose that X and Y have finite expectations. Write

 

Finally, define the population value of squared distance covariance of X and Y as

 

One can show that this is equivalent to the following definition:

 

where E denotes expected value, and     and   are independent and identically distributed. The primed random variables   and   denote independent and identically distributed (iid) copies of the variables   and   and are similarly iid.[5] Distance covariance can be expressed in terms of the classical Pearson's covariance, cov, as follows:

 

This identity shows that the distance covariance is not the same as the covariance of distances, cov(‖XX' ‖, ‖YY' ). This can be zero even if X and Y are not independent.

Alternatively, the distance covariance can be defined as the weighted L2 norm of the distance between the joint characteristic function of the random variables and the product of their marginal characteristic functions:[6]

 

where  ,  , and   are the characteristic functions of (X, Y), X, and Y, respectively, p, q denote the Euclidean dimension of X and Y, and thus of s and t, and cp, cq are constants. The weight function   is chosen to produce a scale equivariant and rotation invariant measure that doesn't go to zero for dependent variables.[6][7] One interpretation of the characteristic function definition is that the variables eisX and eitY are cyclic representations of X and Y with different periods given by s and t, and the expression ϕX, Y(s, t) − ϕX(s) ϕY(t) in the numerator of the characteristic function definition of distance covariance is simply the classical covariance of eisX and eitY. The characteristic function definition clearly shows that dCov2(X, Y) = 0 if and only if X and Y are independent.

Distance variance and distance standard deviation

edit

The distance variance is a special case of distance covariance when the two variables are identical. The population value of distance variance is the square root of

 

where  ,  , and   are independent and identically distributed random variables,   denotes the expected value, and   for function  , e.g.,  .

The sample distance variance is the square root of

 

which is a relative of Corrado Gini's mean difference introduced in 1912 (but Gini did not work with centered distances).[8]

The distance standard deviation is the square root of the distance variance.

Distance correlation

edit

The distance correlation [2][3] of two random variables is obtained by dividing their distance covariance by the product of their distance standard deviations. The distance correlation is the square root of

 

and the sample distance correlation is defined by substituting the sample distance covariance and distance variances for the population coefficients above.

For easy computation of sample distance correlation see the dcor function in the energy package for R.[4]

Properties

edit

Distance correlation

edit
  1.   and  ; this is in contrast to Pearson's correlation, which can be negative.
  2.   if and only if X and Y are independent.
  3.   implies that dimensions of the linear subspaces spanned by X and Y samples respectively are almost surely equal and if we assume that these subspaces are equal, then in this subspace   for some vector A, scalar b, and orthonormal matrix  .

Distance covariance

edit
  1.   and  ;
  2.   for all constant vectors  , scalars  , and orthonormal matrices  .
  3. If the random vectors   and   are independent then
     
    Equality holds if and only if   and   are both constants, or   and   are both constants, or   are mutually independent.
  4.   if and only if X and Y are independent.

This last property is the most important effect of working with centered distances.

The statistic   is a biased estimator of  . Under independence of X and Y [9]

 

An unbiased estimator of   is given by Székely and Rizzo.[10]

Distance variance

edit
  1.   if and only if   almost surely.
  2.   if and only if every sample observation is identical.
  3.   for all constant vectors A, scalars b, and orthonormal matrices  .
  4. If X and Y are independent then  .

Equality holds in (iv) if and only if one of the random variables X or Y is a constant.

Generalization

edit

Distance covariance can be generalized to include powers of Euclidean distance. Define

 

Then for every  ,   and   are independent if and only if  . It is important to note that this characterization does not hold for exponent  ; in this case for bivariate  ,   is a deterministic function of the Pearson correlation.[2] If   and   are   powers of the corresponding distances,  , then   sample distance covariance can be defined as the nonnegative number for which

 

One can extend   to metric-space-valued random variables   and  : If   has law   in a metric space with metric  , then define  ,  , and (provided   is finite, i.e.,   has finite first moment),  . Then if   has law   (in a possibly different metric space with finite first moment), define

 

This is non-negative for all such   iff both metric spaces have negative type.[11] Here, a metric space   has negative type if   is isometric to a subset of a Hilbert space.[12] If both metric spaces have strong negative type, then   iff   are independent.[11]

Alternative definition of distance covariance

edit

The original distance covariance has been defined as the square root of  , rather than the squared coefficient itself.   has the property that it is the energy distance between the joint distribution of   and the product of its marginals. Under this definition, however, the distance variance, rather than the distance standard deviation, is measured in the same units as the   distances.

Alternately, one could define distance covariance to be the square of the energy distance:   In this case, the distance standard deviation of   is measured in the same units as   distance, and there exists an unbiased estimator for the population distance covariance.[10]

Under these alternate definitions, the distance correlation is also defined as the square  , rather than the square root.

Alternative formulation: Brownian covariance

edit

Brownian covariance is motivated by generalization of the notion of covariance to stochastic processes. The square of the covariance of random variables X and Y can be written in the following form:

 

where E denotes the expected value and the prime denotes independent and identically distributed copies. We need the following generalization of this formula. If U(s), V(t) are arbitrary random processes defined for all real s and t then define the U-centered version of X by

 

whenever the subtracted conditional expected value exists and denote by YV the V-centered version of Y.[3][13][14] The (U,V) covariance of (X,Y) is defined as the nonnegative number whose square is

 

whenever the right-hand side is nonnegative and finite. The most important example is when U and V are two-sided independent Brownian motions /Wiener processes with expectation zero and covariance |s| + |t| − |st| = 2 min(s,t) (for nonnegative s, t only). (This is twice the covariance of the standard Wiener process; here the factor 2 simplifies the computations.) In this case the (U,V) covariance is called Brownian covariance and is denoted by

 

There is a surprising coincidence: The Brownian covariance is the same as the distance covariance:

 

and thus Brownian correlation is the same as distance correlation.

On the other hand, if we replace the Brownian motion with the deterministic identity function id then Covid(X,Y) is simply the absolute value of the classical Pearson covariance,

 
edit

Other correlational metrics, including kernel-based correlational metrics (such as the Hilbert-Schmidt Independence Criterion or HSIC) can also detect linear and nonlinear interactions. Both distance correlation and kernel-based metrics can be used in methods such as canonical correlation analysis and independent component analysis to yield stronger statistical power.

See also

edit

Notes

edit

References

edit
edit