Rate of convergence for discretization of integrals with respect to fractional Brownian motion (Q2346985)

The authors consider the rate of the \(L^r\)-convergence for the equidistant discretization of stochastic integrals \(S=\int_{0}^{1}f'_{-}(B_s) dB_s\) with respect to fractional Brownian motion with Hurst parameter \(H>1/2\), where \(f'_{-}\) is the left derivative of a convex function. The stochastic integral, interpreted in generalized Lebesgue-Stieltjes sense and the a.s.\ limit of Riemann sums \(S_n = \sum_{i=1}^{n} f'_{-}(B_{(i-1)/n})(B_{i/n} - B_{(i-1)/n})\), was studied in [\textit{E. Azmoodeh} et al., Stat. Decis. 27, No. 2, 129--143 (2009; Zbl 1202.91312)]. In this article, it is established that \[ \left(\mathbb{E}\left[(S_n-S)^r\right]\right)^{1/r} \leq C \left(\frac{1}{n}\right)^{\frac{H}{p}-\beta} \] for parameters \(p \in (2H,H/(1-H))\), \(r \in [1,p)\), \(\beta \in (1-H,H/p)\) and some constant \(C\). The proof is based on estimates of Besov norms and the structure of the constant \(C\) is specified. Moreover, it is shown that the rate of convergence is arbitrarily close to \(H-1/2\).