Exact and high-order discretization schemes for Wishart processes and their affine extensions (Q1950261): Difference between revisions

The paper is devoted to the simulation of the Wishart processes defined by the following stochastic differential equation (SDE): \[ X_{t}^{x}=x+\int_{0}^{t}(\alpha a^{\intercal}a+bX_{s}^{x}+X_{s}^{x} b^{\intercal})ds+\int_{0}^{t}(\sqrt{X_{s}^{x}}dW_{s}a+a^{\intercal} dW_{s}^{\intercal}\sqrt{X_{s}^{x}}),\tag{1} \] where \(\alpha\geq0\), \(a\), \(b\) are \(d\) square matrices, \(W_{s}\) is a \(d\) square matrix made of independent Brownian motions, \(x\), \(X\) are \(d\) square positive semidefinite matrices. If \(d=1\), (1) is the SDE of the Cox-Ingersoll-Ross process. The authors find a splitting for Wishart processes that makes it possible to sample exactly Wishart distributions. Moreover, they construct high-order discretization schemes for Wishart processes. The authors consider affine extensions of the Wishart processes as well.