Multivariate Pearson diffusions, also known as polynomial diffusions, are characterized by a linear drift vector and a diffusion matrix that is quadratic in the state variables. We derive exact closed-form expressions for the mean and covariance matrix of this class by using results on matrix exponential integrals. We then extend this framework to a broader class of nonlinear diffusions with Pearson-type multiplicative noise. The main contribution of this paper is a new parameter estimator for these nonlinear models based on Strang splitting (SS). The proposed method decomposes the stochastic system into a deterministic nonlinear ordinary differential equation (ODE) and a multivariate Pearson diffusion. We construct the SS estimator by composing their respective flows and applying a Gaussian transition approximation parameterized by the exact moments of the Pearson component. We prove that the SS estimator is consistent and asymptotically efficient. Furthermore, we introduce a new model within this broader class, which we call the Student Kramers oscillator, and we prove existence and uniqueness of the strong solution as well as existence of an invariant measure. We evaluate the SS estimator through simulation studies on this new oscillator and on the multivariate Wright-Fisher diffusion from population genetics. These simulations demonstrate that the SS estimator outperforms the standard Euler-Maruyama estimator, the Gaussian approximation estimator, and the local linearization estimator. Finally, we apply the SS estimator to fit the Student Kramers oscillator to Greenland ice core data.

翻译：暂无翻译