System Reliability Theory: Models and Statistical Method> Marvin Rausand,Arnljot Hoylanc Cowriaht 02004 bv John Wilev & Sons. Inc Appendix C Kronecker Products Let A be a matrix with dimension mA x nA, and B be a matrix with dimension m g x ng. The Kroneckerproduct' of A and B,written A @B is defined as The Kronecker product is also known as the Zehfuss product and the tensor product. A Kronecker sum is an ordinary sum of Kronecker products. The Kronecker sum, written A @B,is defined for square matrices A and B as where n~ is the size of the square matrix A, ne is the size of the square matrix B, and II is the identity matrix. Some Properties of the Kronecker Product 'Named after the GermadPolish mathematician Leopold Kronecker (1823-1891). 581 582 KRONECKER PRODUCTS 1. Associativity: A€3(B€3c)=(A€3B)€3c 2. Distributivity over ordinary matrix addition: (A +B)€3 (e+ D) = A €3 c+ B €3 c+ A €3BD€3 +D 3. Compatibility with ordinary matrix multiplication: AB@cnD= (A@C)(B€3D) 4. Compatibility with ordinary matrix inversion: (A€3 B)-' =A-' @B-' Further details about the Kronecker product may be found in Graham (1981). Kronecker productsare available in many mathematicaltools, like MATLABand Maple. System Reliability Theory: Models and Statistical Method> Marvin Rausand,Arnljot Hoylanc Cowriaht 02004 bv John Wilev & Sons. Inc Appendix D Distribution Theorems First we refer (without proof,)two important theorems from distribution theory: Theorem D.1 Let X be a continuousrandom variablewith probability density f x (x) and sample space S X . Furthermore let a ( x ) be strictly monotonousin x and differentiable with respect to x for all x. Then y = a(x) is a one-to-one transformation on x with inverse x = b ( y ) which maps SX into S y ,and the density of Y = a ( X )is given by fY(Y) (D.1) = f X ( N Y ) ) Ib'(Y)l 0 The extension of this theorem to multivariate distributionsisgiven in the next theorem. Theorem D.2 Let X I ,X2,. ..,X, be continuouslydistributedwithjoint probability density f x ,,x2.....x,, (XI, x2, ...,x,) and sample space S X ,,x2 ,...,X , . If yi = ai(xl,x2,...,x,) for i = 1,2,...,n (D.2) is a one-to-one transformation on the x's with inverse Xj = bj(y1,~ 2 , ..., Y n ) for i = 1.2,. that maps the sample space Sx1,x2,...,X , into Sy, ,y2,...,y,, Yi = ai(X1,X2,... ,X,) for ..,n (D.3) then thejoint density of i = 1,2,...,n 583