A functional limit theorem for self-normalized linear processes with random coefficients and i.i.d. heavy-tailed innovations (Q6054051)

The author considers a linear process of the following form: \(X_{i}=\sum_{j=0}^{\infty }C_{j}Z_{i-j}\), \(i\in\mathbb{Z}\), where \((Z_{i})\) is a strictly stationary sequence of random variables (r.v.) with regularly varying tail \(P(\left\vert Z_{i}\right\vert >x)=x^{-\alpha }L(x)\),\(x>0\) and \(0<\alpha <2\), and \((C_{i})\) is a sequence of r.v. independent of \((Z_{i})\) and so that the above series is a.s. convergent. Let \((a_{n})\) be a sequence of positive numbers such that \(nP(\left\vert Z_{1}\right\vert >a_{n})\rightarrow 1\). In the paper, the author considers the functional convergence of the joint stochastic process \(L_{n}(t)=(a_{n}^{-1}\sum_{i=1}^{\left[ nt\right] }X_{i},a_{n}^{-2} \sum_{i=1}^{\left[ nt\right] }X_{i}^{2})\), \(0\leq t\leq 1\) using the weak Skorohod \(M_{2}\) topology. From this result the author obtains a functional convergence result of the self-normalized sum \(\xi _{n}^{-1}\sum_{i=1}^{\left[ nt \right] }X_{i}\), where \(\xi _{n}^{2}=\sum_{i=1}^{n}X_{i}^{2}\).