A Marcinkiewicz-Zygmund type strong law for weighted sums of \(\phi \)-mixing random variables and its applications (Q2235908)

The paper gives a new generalization of the classical Marcinkiewicz-Zygmund strong law of large numbers, which states that if \(\{ X,{X_n},n \geq 1\} \) is a sequence of independent and identically distributed random variables with \(EX = 0\) and \(E{\left| X \right|^p} < \infty \) for some \(1 \leq p < 2\), then \({n^{ - {1 \mathord{\left/ {\vphantom {1 p}} \right. \kern-\nulldelimiterspace} p}}}\sum\nolimits_{k = 1}^n {{X_k}} \to 0\) almost surely. Its purpose is to establish a strong law for weighted sums \({n^{ - {1 \mathord{\left/ {\vphantom {1 p}} \right. \kern-\nulldelimiterspace} p}}}\sum\nolimits_{k = 1}^{{k_n}} {{a_{nk}}{X_k}} \) of \(\phi \)-mixing random variables without any conditions on mixing rate. The authors impose conditions \({\sup _{n \geq 1}}{{{k_n}} \mathord{\left/ {\vphantom {{{k_n}} n}} \right. \kern-\nulldelimiterspace} n} < \infty \), \({\sup _{n \geq 1}}{\sum\nolimits_{k = 1}^{{k_n}} {\left| {{a_{nk}}} \right|} ^q} < \infty \) for some \(q > 0\), \({\sup _{n \geq 1}}{n^{ - \delta }}\sum\nolimits_{k = 1}^{{k_n}} {a_{nk}^2} < \infty\) for some \(\delta < \min \{ 1,{2 \mathord{\left/ {\vphantom {2 p}} \right. \kern-\nulldelimiterspace} p}\} \). As applications of the main result, they give a convergence rate in the Cesàro law of large numbers for \(\phi \)-mixing random variables, and an almost sure convergence rate for the linear stochastic approximation algorithm.