Limit theorems for logarithmic averages of fractional Brownian motions (Q1970313)

A real-valued process \(\{X(t)\), \(t\geq 0\}\) is called fractional Brownian motion with index \(\alpha, 0<\alpha<1,\) if it is a Gaussian process with \[ EX(t)=0 \quad \text{and} \quad EX(t)X(s)= (1/2)\{t^{2\alpha}+s^{2\alpha}-|t-s|^{2\alpha}\}, \] cf. \textit{G. Samorodnitsky} and \textit{M. S. Taqqu} [Stable non-Gaussian processes (1994)]. Put \(L(T)= \int_1^T(1/t)f(X(t)t^{-\alpha}) dt\). The first result of the authors is an invariance principle for \(L(T)\). Under the supposition \(\int_{-\infty}^\infty|f(x)|^\nu e^{-x^2/2} dx<\infty\) with some \(\nu>2\) there exists a Wiener process \(\{W(t)\), \(t\geq 0\}\) such that \[ |L(T)-\mu\log T-\sigma W(\log T)|=o((\log T)^\varepsilon) \quad \text{a.s.} \] as \(T\to\infty,\) for all \(\varepsilon>\max(1/4,1/\nu)\), where \(\mu=\int_{-\infty}^\infty f(t)\varphi(t) dt\). This yields at once \[ \lim_{T\to\infty}(1/\log T)L(T)= E(f(X(1))\mid \mathcal F) \quad \text{a.s.} \] where \(\mathcal F\) is the tail \(\sigma\)-field of the process \(\{X(t)\), \(t\geq 0\}\), cf. also \textit{G. A. Brosamler} [Math. Proc. Camb. Philos. Soc. 104, No. 3, 561-574 (1988; Zbl 0668.60029)]. As a corollary from their invarince principle, the authors obtain the weak invariance principle and the law of the iterated logarithm for \(L(T)\). Then the authors consider the case when \(f(t)\) is the indicator of short intervals and prove a second invariance principle for \[ Q(T)= \int_1^T(1/t)I\{t^\alpha a(t)\leq X(t)\leq t^\alpha b(t)\}/(\Phi(b(t)) -\Phi(a(t))) dt. \] In the proofs the authors use the stationary Gaussian process \(\{Z(t)=e^{-\alpha t}X(e^t)\), \(-\infty<t<\infty\}\).