Approximate continuous-discrete filters for the estimation of diffusion processes from partial and noisy observations

TL;DR

This paper introduces a new approximation method for estimating parameters of diffusion processes from partial, noisy data, improving accuracy over traditional methods and demonstrating strong performance in simulations.

Contribution

It presents an alternative, convergent approximation to the innovation method using minimum variance filters, enhancing parameter estimation accuracy for diffusion processes.

Findings

New estimators outperform conventional ones in simulations

Estimators are asymptotically normal with decreasing bias

Performance improves as filter approximation error decreases

Abstract

In this paper, an alternative approximation to the innovation method is introduced for the parameter estimation of diffusion processes from partial and noisy observations. This is based on a convergent approximation to the first two conditional moments of the innovation process through approximate continuous-discrete filters of minimum variance. It is shown that, for finite samples, the resulting approximate estimators converge to the exact one when the error of the approximate filters decreases. For an increasing number of observations, the estimators are asymptotically normal distributed and their bias decreases when the above mentioned error does it. A simulation study is provided to illustrate the performance of the new estimators. The results show that, with respect to the conventional approximate estimators, the new ones significantly enhance the parameter estimation of the test equations. The proposed estimators are intended for the recurrent practical situation where a nonlinear stochastic system should be identified from a reduced number of partial and noisy observations distant in time.