Filtered data based estimators for stochastic processes driven by colored noise

We consider the problem of estimating unknown parameters in stochastic differential equations driven by colored noise, given continuous-time observations. Colored noise is modelled as a sequence of mean zero Gaussian stationary processes with an exponential autocorrelation function, with decreasing correlation time. Our goal is to infer parameters in the limit equation, driven by white noise, given observations of the colored noise dynamics. As in the case of parameter estimation for multiscale diffusions, the observations are only compatible with the data in the white noise limit, and classic estimators become biased, implying the need of preprocessing the data. We consider both the maximum likelihood and the stochastic gradient descent in continuous time estimators, and we propose modified versions of these methods, in which the observations are filtered using an exponential filter. Both stochastic differential equations with additive and multiplicative noise are considered. We provide a convergence analysis for our novel estimators in the limit of infinite data, and in the white noise limit, showing that the estimators are asymptotically unbiased. We consider in detail the case of multiplicative colored noise, in particular when the L\'evy area correction drift appears in the limiting white noise equation. A series of numerical experiments corroborates our theoretical results.