Almost Sure Convergence for Linear Processes Generated by Positively Dependent Processes (Q4450719)

The authors obtain strong laws of large numbers for linear processes generated by sequences of random variables, which are either linearly positive quadrant dependent or associated. In this result no stationarity condition is required. The main theorem states as follows: Let \(X_{t}=\sum^{\infty}_{j=0} a_{j} \varepsilon_{t-j}\) be a linear process, where \(\sum^{\infty}_{j=0} | a_{j}| <\infty\) and \(\{\varepsilon_{t}\}\) be a sequence of random variables such that \(E\varepsilon_{t}=0\) and \(0<E \varepsilon^{2}_{t}<\infty\). Assume that \[ \sum^{\infty}_{n=1} n^{-2} \text{Cov}\left(\varepsilon_{n}, \sum^{n}_{t=1} \varepsilon_{t}\right)< \infty, \quad \sup_{t\in N}E| \varepsilon_{t}| ^{s}<\infty, \] for some \(s>2\) and \( u(n)=O(n^{-\rho}), \) for some \(\rho>0.\) Then \( n^{-1}(S_{n} -ES_{n})\rightarrow 0 \text{ a.s.} \) as \(n\rightarrow \infty.\)