On Simpson's rule and fractional Brownian motion with \(H = 1/10\) (Q904713)

In the present paper, the authors study the limiting behaviour of Riemann sums constructed according to Simpson's rule for a fractional Brownian motion. More precisely, let \(B\) be a fractional Brownian motion, that is, a centered Gaussian process with covariance function given by \[ \mathbb{E}[B_s B_t] = \frac{1}{2} \Big( s^{2H} + t^{2H} - |t-s|^{2H} \Big) \quad \text{for all \(s,t \geq 0\),} \] where \(H \in (0,1)\) is the Hurst parameter. Furthermore, let \(f : \mathbb{R} \to \mathbb{R}\) be a smooth function, and consider the Riemann sums constructed according to Simpson's rule with uniform partition: \[ S_n^S(t) := \sum_{j=0}^{\lfloor nt \rfloor - 1} \frac{1}{6} \Big( f' \Big( B_{\frac{j}{n}} \Big) + 4 f' \Big( \Big( B_{\frac{j}{n}} + B_{\frac{j+1}{n}} \Big) / 2 \Big) + f' \Big( B_{\frac{j+1}{n}} \Big) \Big) \Big( B_{\frac{j+1}{n}} - B_{\frac{j}{n}} \Big). \] The authors show that for \(H = 1/10\) this sequence of sums converges weakly to a random variable. More precisely, conditioned on the path \(\{ B_s: s \leq t \}\) one has \[ S_n^S(t) \to f(B_t) - f(0) + \frac{\beta}{2880} \int_0^t f^{(5)}(B_s)\, \mathrm{d}W_s \quad \text{weakly,} \] where \(W\) is a standard Brownian motion, independent of \(B\), and \(\beta \in \mathbb{R}\) is a constant, which is defined in the article. From this result, the authors derive the change-of-variable formula \[ f(B_t) = f(0) + \int_0^t f'(B_s)\, \mathrm{d}^S B_s - \frac{\beta}{2880} \int_0^t f^{(5)}(B_s)\, \mathrm{d}W_s \quad \text{in distribution}, \] where the stochastic integral with differential \(\text{d}^S B_s\) is understood as limit of the Simpson rule sums. The result of the authors contributes to a paper by \textit{M. Gradinaru} et al. [Ann. Inst. Henri Poincaré, Probab. Stat. 41, No. 4, 781--806 (2005; Zbl 1083.60045)], where it was shown that for \(H > 1/10\) the sequence of sums converges in probability, but generally does not converge in probability for \(H \leq 1/10\).