Correlated continuous time random walks (Q1017816)

Let \((Z_j: j\in \mathbb{Z})\) and \((J_k: k\in \mathbb{N})\), \(J_k>0\), be independent sequences of i.i.d. random variables, and \((N_t: t\geq 0)\) be a renewal process generated by the \(J_k\)'s. Assume that the laws of \(Z_1\) and \(J_1\) belong to the domains of attraction of a strictly \(\alpha\)-stable law, \(\alpha\in (0,2]\), and of a \(\beta\)-stable law, \(\beta\in (0,1)\), respectively. Define \(Y_n:=\sum_{j=0}^\infty c_j Z_{n-j}\), \(n\in \mathbb{Z}\) for some real constants \(c_j\) satisfying certain conditions which, among others, ensure the a.s. convergence of the latter series. Depending on summability properties of \(c_j\)'s, the paper under review proves four weak convergence results for, properly scaled, continuous time random walk \((Y_1+\ldots+Y_{N_t}: t\geq 0)\). In particular, it is shown that the set of possible limiting processes includes four distinct processes subordinated to an inverse \(\beta\)-stable subordinator (1) a stable subordinator (convergence holds in the \(M_1\) topology on \(D[0,\infty)\)); (2) a linear fractional stable motion (LFSM) with a.s. continuous paths (convergence holds in the \(J_1\) topology on \(D[0,\infty)\)); (3) a LFSM with a.s. unbounded paths on every interval of positive length (convergence of finite-dimensional distributions); (4) a fractional Brownian motion (convergence holds in the \(J_1\) topology on \(D[0,\infty)\)). Notice that the first case bears resemblance with a classical situation when \((Y_n: n\in \mathbb{Z})\) are i.i.d.