English

Nonparametric statistical inference for drift vector fields of multi-dimensional diffusions

Statistics Theory 2020-07-21 v3 Numerical Analysis Probability Statistics Theory

Abstract

The problem of determining a periodic Lipschitz vector field b=(b1,,bd)b=(b_1, \dots, b_d) from an observed trajectory of the solution (Xt:0tT)(X_t: 0 \le t \le T) of the multi-dimensional stochastic differential equation \begin{equation*} dX_t = b(X_t)dt + dW_t, \quad t \geq 0, \end{equation*} where WtW_t is a standard dd-dimensional Brownian motion, is considered. Convergence rates of a penalised least squares estimator, which equals the maximum a posteriori (MAP) estimate corresponding to a high-dimensional Gaussian product prior, are derived. These results are deduced from corresponding contraction rates for the associated posterior distributions. The rates obtained are optimal up to log-factors in L2L^2-loss in any dimension, and also for supremum norm loss when d4d \le 4. Further, when d3d \le 3, nonparametric Bernstein-von Mises theorems are proved for the posterior distributions of bb. From this we deduce functional central limit theorems for the implied estimators of the invariant measure μb\mu_b. The limiting Gaussian process distributions have a covariance structure that is asymptotically optimal from an information-theoretic point of view.

Keywords

Cite

@article{arxiv.1810.01702,
  title  = {Nonparametric statistical inference for drift vector fields of multi-dimensional diffusions},
  author = {Richard Nickl and Kolyan Ray},
  journal= {arXiv preprint arXiv:1810.01702},
  year   = {2020}
}

Comments

55 pages, to appear in the Annals of Statistics

R2 v1 2026-06-23T04:27:05.445Z