Fourier series approximation of linear fractional stable motion (Q2483006)

Irregular phenomena appear in various fields of scientific research as traffic volume in modern communications, computer network, mathematical finance and others. Mathematical models which are both easy to use and relevant for these applications, are fractional stable motions, most prominently of which is the fractional Brownian motion. In the introduction for \(0 < \alpha \leq 2\), the concept of an independently scattered symmetric \(\alpha\)-stable is given. The process \(\{X(t)\}_{t \in {\mathbb R}},\) called linear fractional stable motion (LFSM) is introduced. The so-called well balanced linear fractional stable motion is introduced, too. The purpose of the present paper is to present a Fourier series approximation of the LSFM \(\{X(t)\}_{t \in {\mathbb R}},\) which is used to efficiently simulate its sample paths. In Section 2, the general idea of the used approximation method for LFSM in the case \(0 < H < 1\), and \(1 < \alpha \leq 2\) using Fourier series is described. The quantities \(Y_{A}(t),\) \(Y_{A,M}(t)\) and \(Y_{A,M,L}(t)\) for \(t \in {\mathbb R}\) are defined. In Theorem 1 for \(0 < \alpha \leq 2\), the convergence of \(Y_{A,M,L}(t)\) to \(Y_{A,M}(t)\) when \(L \to \infty,\) of \(Y_{A,M}(t)\) to \(Y_{A}(t)\) when \(M \to \infty\) and \(Y_{A}(t)\) to \(X(t)\) when \(A \to \infty\) in probability for any fixed \(t \in {\mathbb R}\) is confirmed. In Section 3, error estimates of the approximations \(Y_{A,M,L}(t)_{t \in {\mathbb R}}\) to \(Y_{A}(t)_{t \in {\mathbb R}}\) and latter to the LFSM \(\{X(t)\}_{t \in {\mathbb R}}\) in terms of the scale parameter are given. In Proposition 1, the effect of truncation on the integral defining \(X(t)\) is analyzed. In Theorem 2 for \(1 < \alpha \leq 2\) and \(1/ \alpha < H < 1\), a common estimation of the \(\alpha\)-norm of \(Y_{A}(t) - Y_{A,M,L}(t)\) in two cases is presented. As consequences of Theorem 2, two Corollaries are obtained. In Corollary 2, the convergence of all finite-dimensional marginal distributions is obtained. In Section 4 an effective simulation algorithm of the sample path of the LFSM \(\{X(t)\}_{t \in {\mathbb R}}\) is presented. Concrete realizations of LFSM for different \(H\) and \(\alpha\) are illustrated. The computational times are compared. In the appendix the proofs of Proposition 1 and Theorem 2 are presented.