On the weak invariance principle for non-adapted sequences under projective criteria (Q2471130)

Let \((\Omega,{\mathcal A},P)\) be a probability space and \(T: \Omega\to\Omega\) be a bijective bimeasurable transformation preserving the probability \(P\). Let \(X_0\) be a square integrable random variable with mean \(0\) and define the stationary sequence \((X_i)_{i\in\mathbb{Z}}\) by \(X_i= X_0\circ T^i\). We denote by \({\mathcal T}\) the \(\sigma\)-algebra of all \(T\)-invariant sets. For a \(\sigma\)-algebra \({\mathcal M}_0\) satisfying \({\mathcal M}_0\subseteq T^{-1}({\mathcal M}_0)\) we define the nondecreasing filtration \(({\mathcal M}_i)_{i\in\mathbb{Z}}\) by \({\mathcal M}_i= T^{-1}({\mathcal M}_0)\). Let \(H_i\) be the space of \({\mathcal M}_i\)-measurable and square integrable random variables and \(m\) an element of \(H_0\) orthogonal to \(H_{-1}\). Let \(S_n X_1+\cdots+ X_n\) and \((s_n)_{n\geq 1}\) a sequence of positive numbers such that \(s_n\to\infty\). In this paper, the authors obtain, among others, the following results: (I) Assume that \(s_{[nt]}/s_n\) is bounded for any \(t\in[0, 1]\). If conditions \(C_1(s_n)(b)\) \[ \lim_{n\to\infty} \sum^n_{l=1} \Biggl\|{\sqrt{n}\over s_n} \sum^{n-l}_{i= 1-l} P_0(X_i)- m\Biggr\|^2_2= 0 \] and \(C_4(s_n),\) (a) \(\| \sup_{1\leq k\leq n}|E(S_k|{\mathcal M}_0\|\,|_2= o(s_n)\), \(\|\sup_{1\leq k\leq n}|S_k- E(S_k|{\mathcal M}_n)|\,\|_2= o(s_n)\), (b) for some positive sequence \((u_i)_{i\in\mathbb{Z}}\) such that \((\sqrt{n}/s_n)u_i\) is bounded, \[ \lim_{A\to\infty}\,\limsup_{n\to\infty}\,{\sqrt{n}\over s_n} \sum^n_{i=-n} E\Biggl({P^2_0(X_i)\over u_i} I_{P^2_0(X_i)> Au^2_i}\Biggr)= 0 \] hold, then \(\{s^{-1}_n S_{[nt]}, t\in [0,1]\}\) converges in distribution to \(\sqrt{E(m^2|{\mathcal T})}W\), where \(W\) is a standard Brownian motion independent of \({\mathcal T}\) (Theorem 2). (II) In particular, if condition \(C_3\) \[ X_0\text{ is regular and }\sum_{i\in\mathbb{Z}}\| P_0(X_i)\|_2<\infty \] holds, the assertion in (I) is valid with \(E(m^2|{\mathcal T})= \sum_{k\in\mathbb{Z}} E(X_0 X_k|{\mathcal T})\). (Corollary 2). Using the result they establish weak invariance principles for linear processes with dependent innovations and regular functions of linear processes, etc.