Limit theorems for functionals of long memory linear processes with infinite variance (Q6145598): Difference between revisions

Let \ \(X=\left\{ X_{n},n\in\mathbb{N}\right\} \) be a linear process defined by \(X_{n}=\sum_{i=1}^{\infty}a_{i}\varepsilon _{n-i}\), where the innovations \(\varepsilon _{i}\) are i.i.d. and are in the domain of attraction of an \(\alpha \)-stable law with \(0<\alpha <2\), \(\varepsilon _{i}\) is centered for \(\alpha >1\) and symmetric for \(\alpha =1\). The authors assume that \(a_{i}\) is regularly varying with index \(\beta \) with \(\alpha \beta >1\). For an integrable and square integrable function \(K(.)\) the authors study the asymptotic behaviour of the partial sums process \(\left\{ S_{\left[ Nt\right] }=\sum_{n=1}^{\left[ Nt \right] }(K(X_{n})-EK(X_{n})\right\} \) as \(N\rightarrow \infty \). In the main result of the paper the authors establish that the finite-dimensional distributions of \(\left\{ A(N)S_{\left[ Nt\right] }\right\} \) converge to those of another (explicit known) process. The normalizing sequence \(A(n)\) depends on all ingredients mentioned above. The proof of the result is lengthy and rather technical.