Trajectory fitting estimators for SPDEs driven by additive noise (Q1744219): Difference between revisions

The authors study a parameter estimation problem for the drift/viscosity coefficient of a linear parabolic diagonalizable stochastic partial differential equation driven by an additive space-time noise. The observable variable is one path of the solution, observed continuously on a finite time-interval. They propose a new type of estimators, related to what is konwn as the trajectory fitting estimators (TFEs). Using as observations the first \(N\) Fourier modes, they construct the TFE by analogy to the TFE for continuously observed finite dimensional ergodic diffusion processes or similar to the least squares estimators used in time-series analysis. Using the diagonalizable nature of the equations, they get explicit expressions for the estimators and for the asymptotics of the their first two moments. Under some assumptions, they prove consistency and asymptotic normality of the estimators.