Nonparametric Bayesian posterior contraction rates for discretely observed scalar diffusions (Q2403433)

This paper considers the nonparametric Bayesian inference for the drift function \(b\) and the diffusion coefficient \(\sigma\) of the diffusion model, \(dX_t=b(X_t) dt + \sigma(X_t)dW_t\), where \(W_t\) is a standard Brownian motion, with discretely sampled observations \(X_0, X_{\Delta}, \ldots, X_{n\Delta}\). The authors prove the main theorem that characterizes the nonparametric prior distributions on the diffusion parameters \((\sigma, b)\) giving rise to posterior distributions that contract at the optimal convergence rates. The two key ingredients for the proof are the characterization of the small ball probability measured by the information-theoretic distance, and the construction of a sequence of tests for estimators of \((\sigma, b)\). The authors also provide an example of a class of prior distributions where the main theorem applies, which is defined by random wavelet series.