Matrix variate beta distribution

From HandWiki - Reading time: 3 min

In statistics, the matrix variate beta distribution is a generalization of the beta distribution. If U is a p×p positive definite matrix with a matrix variate beta distribution, and a,b>(p1)/2 are real parameters, we write UBp(a,b) (sometimes BpI(a,b)). The probability density function for U is:

{βp(a,b)}1det(U)a(p+1)/2det(IpU)b(p+1)/2.


Matrix variate beta distribution
Notation Bp(a,b)
Parameters a,b
Support p×p matrices with both U and IpU positive definite
PDF {βp(a,b)}1det(U)a(p+1)/2det(IpU)b(p+1)/2.
CDF 1F1(a;a+b;iZ)

Here βp(a,b) is the multivariate beta function:

βp(a,b)=Γp(a)Γp(b)Γp(a+b)

where Γp(a) is the multivariate gamma function given by

Γp(a)=πp(p1)/4i=1pΓ(a(i1)/2).

Theorems

Distribution of matrix inverse

If UBp(a,b) then the density of X=U1 is given by

1βp(a,b)det(X)(a+b)det(XIp)b(p+1)/2

provided that X>Ip and a,b>(p1)/2.

Orthogonal transform

If UBp(a,b) and H is a constant p×p orthogonal matrix, then HUHTB(a,b).

Also, if H is a random orthogonal p×p matrix which is independent of U, then HUHTBp(a,b), distributed independently of H.

If A is any constant q×p, qp matrix of rank q, then AUAT has a generalized matrix variate beta distribution, specifically AUATGBq(a,b;AAT,0).

Partitioned matrix results

If UBp(a,b) and we partition U as

U=[U11U12U21U22]

where U11 is p1×p1 and U22 is p2×p2, then defining the Schur complement U221 as U22U21U111U12 gives the following results:

  • U11 is independent of U221
  • U11Bp1(a,b)
  • U221Bp2(ap1/2,b)
  • U21U11,U221 has an inverted matrix variate t distribution, specifically U21U11,U221ITp2,p1(2bp+1,0,Ip2U221,U11(Ip1U11)).

Wishart results

Mitra proves the following theorem which illustrates a useful property of the matrix variate beta distribution. Suppose S1,S2 are independent Wishart p×p matrices S1Wp(n1,Σ),S2Wp(n2,Σ). Assume that Σ is positive definite and that n1+n2p. If

U=S1/2S1(S1/2)T,

where S=S1+S2, then U has a matrix variate beta distribution Bp(n1/2,n2/2). In particular, U is independent of Σ.

See also

References

  • Gupta, A. K.; Nagar, D. K. (1999). Matrix Variate Distributions. Chapman and Hall. ISBN 1-58488-046-5. 
  • Mitra, S. K. (1970). "A density-free approach to matrix variate beta distribution". The Indian Journal of Statistics. Series A (1961–2002) 32 (1): 81–88. 




Licensed under CC BY-SA 3.0 | Source: https://handwiki.org/wiki/Matrix_variate_beta_distribution
6 views | Status: cached on August 12 2024 11:45:59
↧ Download this article as ZWI file
Encyclosphere.org EncycloReader is supported by the EncyclosphereKSF