Complex Wishart distribution explained
In statistics, the complex Wishart distribution is a complex version of the Wishart distribution. It is the distribution of
times the sample Hermitian covariance matrix of
zero-mean
independent Gaussian random variables. It has
support for
Hermitian positive definite matrices.
[1] The complex Wishart distribution is the density of a complex-valued sample covariance matrix. Let
where each
is an independent column
p-vector of random complex Gaussian zero-mean samples and
is an Hermitian (complex conjugate) transpose. If the covariance of
G is
then
where
is the complex central Wishart distribution with
n degrees of freedom and mean value, or scale matrix,
M.
fS(S)=
|
\left|S\right|n-pe-\operatorname{tr(M-1S) |
}
{
\left|M\right| |
n ⋅ l{C}\widetilde{\Gamma}p(n)
}, n\gep, \left|M\right|>0
where
(n)=\pip(p-1)/2
\Gamma(n-j+1)
is the complex multivariate Gamma function.
[2] Using the trace rotation rule
\operatorname{tr}(ABC)=\operatorname{tr}(CAB)
we also get
fS(S)=
|
\left|S\right|n-p
|
\left|M\right|n ⋅ l{C |
\widetilde{\Gamma}p(n)}\exp\left(
-1Gi\right)
which is quite close to the complex multivariate pdf of
G itself. The elements of
G conventionally have circular symmetry such that
.
Inverse Complex WishartThe distribution of the inverse complex Wishart distribution of
according to Goodman,
[3] Shaman
[4] is
fY(Y)=
|
\left | |Y\right|-(n+p)e-\operatorname{tr |
|
}
{
\left|M\right| |
-n ⋅ l{C}\widetilde{\Gamma}p(n)
}, n\gep, \det\left(Y\right)>0
where
.
If derived via a matrix inversion mapping, the result depends on the complex Jacobian determinant
Goodman and others[5] discuss such complex Jacobians.
Eigenvalues
The probability distribution of the eigenvalues of the complex Hermitian Wishart distribution are given by, for example, James[6] and Edelman.[7] For a
matrix with
degrees of freedom we have
f(λ1...λp)=\tilde{K}\nu,p\exp\left(-
λi\right)
\prodi<j(λi-
dλ1...dλp,
λi\inR\ge0
where
\tilde
=2p\nu
\Gamma(\nu-i+1)\Gamma(p-i+1)
Note however that Edelman uses the "mathematical" definition of a complex normal variable
where iid
X and
Y each have unit variance and the variance of
. For the definition more common in engineering circles, with
X and
Y each having 0.5 variance, the eigenvalues are reduced by a factor of 2.
While this expression gives little insight, there are approximations for marginal eigenvalue distributions. From Edelman we have that if S is a sample from the complex Wishart distribution with
p=\kappa\nu, 0\le\kappa\le1
such that
Sp\siml{CW}\left(2I,
\right)
then in the limit
the distribution of eigenvalues converges in probability to the
Marchenko–Pastur distribution function
pλ(λ)=
| \sqrt{[λ/2-(\sqrt{\kappa |
-1 |
)2][\sqrt{\kappa}+1)2-λ/2]}}
{4\pi\kappa(λ/2)}, 2(\sqrt{\kappa}-1)2\leλ\le2(\sqrt{\kappa}+1)2,
0\le\kappa\le1
This distribution becomes identical to the real Wishart case, by replacing
by
, on account of the doubled sample variance, so in the case
Sp\siml{CW}\left(I,
\right)
, the pdf reduces to the real Wishart one:
pλ(λ)=
)2][\sqrt{\kappa}+1)2-λ]}}
{2\pi\kappaλ}, (\sqrt{\kappa}-1)2\leλ\le(\sqrt{\kappa}+1)2,
0\le\kappa\le1
A special case is
pλ(λ)=
\left(
\right
, 0\leλ\le8
or, if a Var(
Z) = 1 convention is used then
pλ(λ)=
\left(
\right
, 0\leλ\le4
.The
Wigner semicircle distribution arises by making the change of variable
in the latter and selecting the sign of
y randomly yielding pdf
py(y)=
\left(4-y2\right
, -2\ley\le2
In place of the definition of the Wishart sample matrix above,
, we can define a Gaussian ensemble
such that
S is the matrix product
. The real non-negative eigenvalues of
S are then the modulus-squared singular values of the ensemble
and the moduli of the latter have a quarter-circle distribution.
In the case
such that
then
is rank deficient with at least
null eigenvalues. However the singular values of
are invariant under transposition so, redefining
, then
has a complex Wishart distribution, has full rank almost certainly, and eigenvalue distributions can be obtained from
in lieu, using all the previous equations.
In cases where the columns of
are not linearly independent and
remains singular, a
QR decomposition can be used to reduce
G to a product like
G=Q\begin{bmatrix}R\ 0\end{bmatrix}
such that
is upper triangular with full rank and
has further reduced dimensionality.
The eigenvalues are of practical significance in radio communications theory since they define the Shannon channel capacity of a
MIMO wireless channel which, to first approximation, is modeled as a zero-mean complex Gaussian ensemble.
Notes and References
- N. R. Goodman . 1963 . The distribution of the determinant of a complex Wishart distributed matrix . The Annals of Mathematical Statistics . 34 . 1 . 178–180. 10.1214/aoms/1177704251 . free .
- Goodman. N R. 1963. Statistical analysis based on a certain multivariate complex Gaussian distribution (an introduction). Ann. Math. Statist.. 34. 152–177. 10.1214/aoms/1177704250. free.
- Goodman. N R. 1963. Statistical analysis based on a certain multivariate complex Gaussian distribution (an introduction). Ann. Math. Statist.. 34. 152–177. 10.1214/aoms/1177704250. free.
- Shaman. Paul. 1980. The Inverted Complex Wishart Distribution and Its Application to Spectral Estimation. Journal of Multivariate Analysis. 10. 51–59. 10.1016/0047-259X(80)90081-0. free.
- Web site: On the Relation between Real and Complex Jacobian Determinants. Cross. D J. May 2008. drexel.edu.
- James. A. T.. 1964. Distributions of Matrix Variates and Latent Roots Derived from Normal Samples. Ann. Math. Statist.. 35. 2. 475–501. 10.1214/aoms/1177703550. free.
- Edelman. Alan. October 1988. Eigenvalues and Condition Numbers of Random Matrices. SIAM J. Matrix Anal. Appl.. 9 . 4. 543–560. 10.1137/0609045. 1721.1/14322. free.