Convergence to Lévy stable processes under some weak dependence conditions (Q988675): Difference between revisions

Let \(\{Z_j\}\) be a strictly stationary, strongly mixing sequence of \(\mathbb{R}^d\)-valued random vectors with rate function \(\phi_0\). We assume that each \(Z_j\) is regularly varying with index \(\alpha\) in \((0,2)\). Choose sequences \(b_n\) and \(c_n\) such that \[ nP(|Z_1|> b_n)= 1\text{ and }c_n= {n\over b_n} E(Z_1I(|Z_1|\leq b_n)) \] and define the partial sum processes \[ X_n(t)={1\over b_n}\,\sum_{1\leq j\leq nt} Z_j- tc_n,\quad t\geq 0. \] In this paper, the author proves that \(X_n\to^dX\) in \(D([0,\infty), \mathbb{R}^d)\), where \(X\) is a Lévy \(\alpha\)-stable process, if and only if for any \(\varepsilon> 0\) there exist sequences of integers \(r_n(\varepsilon)\) and \(\ell_n(\varepsilon)\) such that \(r_n= o(n)\), \(\ell_n= o(r_n)\), \(n\phi_0(\ell_n)= o(r_n)\) and \[ \lim_{n\to\infty}\,P\Biggl(\max_{2\leq j\leq r_n}|Z_j|>\varepsilon b_n\,\Biggl|\,|Z_1|>\varepsilon b_n\Biggr)= 0 \] assuming that one of the following conditions holds: (1) \(\alpha\in(0, 1)\); (2) \(\alpha\in[1, 2)\) and for every \(\delta> 0\) \[ \lim_{\varepsilon\to 0}\,\limsup_{n\to \infty}\, P\Biggl(\max_{1\leq k\leq n}\,\Biggl|\,\sum^k_{j=1} \Biggl(Z_j I(|Z_j|\leq\varepsilon b_n)- EZ_1I(|Z_1|\leq\varepsilon b_n\Biggr)\Biggr|\geq\delta b_n\Biggr)= 0. \] To establish this theorem, the author proves, under more general settings than above, a theorem which describes the connection between convergence in the Skorohod space with \(J_1\)-topology and convergence of the corresponding point processes in \(M_p((0, \infty)\times\overline{\mathbb{R}}^d_0)\).