Invariance principles and central limit theorems for nonadditive, stationary processes (Q4026052)

[For the entire collection see Zbl 0744.00035.] Let \({\mathcal X}=(X_{m,n})_{0\leq m\leq n<\infty}\) be a superadditive (or subadditive) process stationary in the sense that \((X_{m,n}){\buildrel{\text{d}} \over =}(X_{m+1,n+1})\). The author considers an asymptotic behaviour of variance \(V(X_{0,N})\) and proves that if \({\mathcal X}\) is an independent superadditive process which satisfies central limit theorem and the condition (i) there exists an aditive process \({\mathcal Y}=(Y_{m,n})_{0\leq m\leq n<\infty}\) with \(\| X_{0,N}-Y_{0,N}\|_{L_ 1}=o\left(\sqrt{V(X_{0,N})}\right)\) such that the joint process \((X_{m,n},Y_{m,n})_{0\leq m\leq n<\infty}\) is stationary, then (ii) \(V(X_{0,N})=Nh(N)\) where \(h\) is a slowly varying function. Under slightly generalized \(\alpha\)-mixing condition the central limit theorem is proved for stationary nonadditive process \({\mathcal X}\) satisfying (i), (ii) and the condition (iii) the sequence of random variables \(\{V(X_{0,N})^{-1}(X_{0,N}- EX_{0,N})^ 2\}_{N\in\mathbb{N}}\) is uniformly integrable. Several results concerning a verification of conditions (ii) and (iii) for various mixing conditions are obtained. In the second part of the paper the author obtains an almost sure approximation of the redefined process \({\mathcal X}\) by standard Brownian motion. This is achieved under the following assumptions: (iv) \({\mathcal X}\subset L_{2+\delta}\) \((0<\delta\leq 1)\) satisfies the condition (i) and a generalized \(\alpha\)-mixing condition with the mixing rate \(O(N^{-x})\) for some \(x>2\), (v) there exist constants \(0<\varepsilon<1/10\) and \(0<\sigma<\infty\) such that \(V(X_{0,N})=\sigma^ 2N+O(N^{1-\varepsilon})\) and \(\| X_{0,N}-Y_{0,N}\|_{L_ 1}=o(N^{1/2-4\varepsilon})\), (vi) there exist nonnegative constants \(\alpha,\theta\) and \(\rho\) satisfying some additional inequalities such that \[ \sup_ k\bigl(\sup_{N\geq k^ \rho}\bigl(N^{-1-\alpha/2-\theta}\|\hat X_{0,N+k}-\hat X_{0,k}\|^{2+\alpha}_{L_{2+\alpha}}\bigr)\bigr)<\infty, \] where \(\hat X=X-EX\). Under the same assumptions two almost sure invariance principles are obtained, replacing (vi) by conditions which characterize the closeness of processes \({\mathcal X}\) and \({\mathcal Y}\) with respect to the norm of spaces \(L_ p\) \((p\geq 2)\).