A comparative linear mean-square stability analysis of Maruyama- and Milstein-type methods (Q632730): Difference between revisions

The authors consider the scalar linear stochastic differential equation \[ dX(t) = \lambda X(t)dt + \sum_{r=1}^m \mu_r X(t)dW_r(t), \quad t \geq t_0 \geq 0, \quad X(t_0)= X_0, \] driven by an \(m\)-dimensional standard Wiener process \(W(t) = (W_1(t), \dots, W_m(t))\). They then compare the mean-square stability properties of the \(\theta\)-Maruyama and the \(\theta\)-Milstein method when applied to this test equation. In particular, they provide necessary and sufficient conditions for mean-square stability in each case, showing that the condition in the case of the \(\theta\)-Milstein method involves terms in addition to those also present for the \(\theta\)-Maruyama method, which explicitly depend on the coefficients of the diffusion term. Further, by introducing a method parameter \(\sigma\) into the diffusion approximation terms of the Milstein-type methods, thus obtaining a partial implicitness in these diffusion approximation terms, they study the effect of having some control over the stability properties of these methods. Numerical examples illustrate the results and provide a comparison of the stability behaviour of the different methods.