On tightness of partial sums from strictly stationary, absolutely regular sequences of \(B\)-valued random variables (Q1963969)

Let \(B\) be a real Banach space. Let \(X= \{X_k\}\) be a strictly stationary sequence of \(B\)-valued random variables defined on a probability space \((\Omega,{\mathcal F}, P)\). Suppose \(X\) satisfies the absolute regularity condition that \[ \beta(n)= \sup{1\over 2} \sum^I_{i= 1} \sum^J_{j= 1}|P(G_i\cap H_j)- P(G_i) P(H_j)|\to 0\qquad (n\to\infty), \] where the supremum is taken over all pairs of finite partitions \(\{G_1,\dots, G_I\}\) and \(\{H_1,\dots, H_J\}\) of \(\Omega\) such that \(G_i\in{\mathcal F}^0_{-\infty}\) for all \(i\) and \(H_j\in{\mathcal F}^\infty_n\) for all \(j\) and \({\mathcal F}^b_a\) denotes the \(\sigma\)-algebra generated by \(X_a,\dots, X_b\). The author shows that the property of tightness of the partial sums of \(X\) follows from shift-tightness of the bounded linear functionals of those sums. More precisely, the following are shown. Let \(S_n= X_1+\cdots+ X_n\). (1) The sequence \(\{S_n\}\) satisfies either shift-tightness or complete dissipation (in \(\mathbb{R}\)). (2) If there exists a real bounded linear functional \(f\) on \(B\) such that the sequence \(\{f(S_n)\}\) fails to satisfy shift-tightness (in \(\mathbb{R}\)), then \(\{S_n\}\) satisfies complete dissipation (in \(B\)). (3) If for every real bounded linear functional \(f\) on \(B\), \(\{f(S_n)\}\) satisfies shift-tightness (in \(\mathbb{R}\)), then \(\{S_n\}\) satisfies shift-tightness (in \(B\)). (4) If for every real bounded linear functional \(f\) on \(B\), \(\{f(S_n)\}\) satisfies shift-tightness (in \(\mathbb{R}\)), then \(\{S_n\}\) satisfies tightness (in \(B\)) and there exists a strictly stationary, absolutely regular sequence \(\{Z_k\}\) of \(B\)-valued random variables such that for all \(k\), \(X_k= Z_k- Z_{k- 1}\) a.s.