A ridge estimator of the drift from discrete repeated observations of the solution of a stochastic differential equation (Q1983630)

This paper introduces the estimation of the drift of a diffusion equation in the framework of functional data analysis. The observations are \(N\) independent replicates of high-frequency discretized solutions of a diffusion equation on a fixed time interval \([0,T]\). The estimator of the drift is constructed with the \(B\)-splines basis, the coefficients are determined through a least square technique with a ridge \(\ell^2\) penalty. Since the estimation is based on \(N\) replicates, the properties of such an estimator are studied under the asymptotics \(N \rightarrow +\infty\). In particular, consistency and rate of convergence are analyzed. The simulation study shows the practical applicability of the proposed estimator in several scenarios and proves that the ridge estimator outperforms the estimator based on the cut-off procedure from previous literature.