Nonparametric estimation for stochastic volatility models (Q2430253)

The main object of the paper is a stochastic volatility model, where the unobserved volatility process satisfies a certain time-homogeneous stochastic differential equation. The authors give a new nonparametric method for estimating both the drift and the diffusion coefficient of this equation, based on a finite number of observations taken at equidistant points. This method is based on a least squares approach, where the estimators are chosen among a collection of functions belonging to a finite-dimensional space whose dimension is selected by a data driven procedure. Bounds of the risk are given, and the method proposed by the authors provides adaptive estimators that can be implemented by feasible algorithms. The authors give an example of such algorithms together with numerical results, which illustrate the efficiency of their approach.