| A kernel-type estimator is studied for a class of stochastic differential equations with time-dependent coefficients. Both ordinary differential equation and equations with partial derivatives are considered. While the existing theoretical results guarantee the mean-square convergence of the estimate as the number of observations increases, such results provide no information about estimator's performance on a particular finite set of observations. With the help of computer simulations, it is shown in the current work that individual realizations of the estimator are sensitive to the initial condition of the observation process, and, to a certain point, the performance of the estimator is improved by increasing the initial condition. In the case of the stochastic heat equation, the performance of the estimator can also be improved by reducing the eigenvalues of the corresponding elliptic operator. |
PDF Full Text Request