Matrix F-distribution

Matrix F
Notation ℱ(𝜳,ν,δ)
Parameters 𝜳>0, p×p scale matrix (pos. def.)
ν>p−1 degrees of freedom (real)
δ>0 degrees of freedom (real)
Support 𝐗 is p × p positive definite matrix
PDF

Γp(ν+δ+p−12)Γp(ν2)Γp(δ+p−12)|𝜳|ν2 |𝐗|ν−p−12|𝐈p+𝐗𝜳−1|−ν+δ+p−12

Mean νδ−2𝜳, for δ>2.
Variance see below

In statistics, the matrix F distribution (or matrix variate F distribution) is a matrix variate generalization of the F distribution which is defined on real-valued positive-definite matrices. In Bayesian statistics it can be used as the semi conjugate prior for the covariance matrix or precision matrix of multivariate normal distributions, and related distributions.

Density

The probability density function of the matrix F distribution is:

f𝐗(𝐗;𝜳,ν,δ)=Γp(ν+δ+p−12)Γp(ν2)Γp(δ+p−12)|𝜳|ν2 |𝐗|ν−p−12|𝐈p+𝐗𝜳−1|−ν+δ+p−12

where 𝐗 and 𝜳 are p×p positive definite matrices, |⋅| is the determinant, Γp(⋅) is the multivariate gamma function, and 𝐈p is the p × p identity matrix.

Properties

Construction of the distribution

  • The standard matrix F distribution, with an identity scale matrix 𝐈p, was originally derived by. When considering independent distributions,

𝜱1∼𝒲(𝐈p,ν) and 𝜱2∼𝒲(𝐈p,δ+k−1), and define 𝐗=𝜱2−1/2𝜱1𝜱2−1/2, then 𝐗∼ℱ(𝐈p,ν,δ).

  • If 𝐗|𝜱∼𝒲−1(𝜱,δ+p−1) and 𝜱∼𝒲(𝜳,ν), then, after integrating out 𝜱, 𝐗 has a matrix F-distribution, i.e.,

f𝐗|𝜱,ν,δ(𝐗)=∫f𝐗|𝜱,δ+p−1(𝐗)f𝜱|𝜳,ν(𝜱)d𝜱.
This construction is useful to construct a semi-conjugate prior for a covariance matrix.

  • If 𝐗|𝜱∼𝒲(𝜱,ν) and 𝜱∼𝒲−1(𝜳,δ+p−1), then, after integrating out 𝜱, 𝐗 has a matrix F-distribution, i.e.,
    f𝐗|𝜳,ν,δ(𝐗)=∫f𝐗|𝜱,ν(𝐗)f𝜱|𝜳,δ+p−1(𝜱)d𝜱.
    This construction is useful to construct a semi-conjugate prior for a precision matrix.

Marginal distributions from a matrix F distributed matrix

Suppose 𝐀∼F(𝜳,ν,δ) has a matrix F distribution. Partition the matrices 𝐀 and 𝜳 conformably with each other

𝐀=[𝐀11𝐀12𝐀21𝐀22],𝜳=[𝜳11𝜳12𝜳21𝜳22]

where 𝐀ij and 𝜳ij are pi×pj matrices, then we have 𝐀11∼F(𝜳11,ν,δ).

Moments

Let X∼F(𝜳,ν,δ).

The mean is given by: E(𝐗)=νδ−2𝜳.

The (co)variance of elements of 𝐗 are given by:

cov⁡(Xij,Xml)=ΨijΨml2ν2+2ν(δ−2)(δ−1)(δ−2)2(δ−4)+(ΨilΨjm+ΨimΨjl)(2ν+ν2(δ−2)+ν(δ−2)(δ−1)(δ−2)2(δ−4)+ν(δ−2)2).

See also