Power series methods and almost sure convergence

From MaRDI portal





The power series \(p(t)=\sum^ \infty_{k=0}p_ kt^ k\) with \(p_ 0>0\), \(p_ k\geq 0\), \(k\geq 1\), and convergence radius \(R\) provides the summability method P by defining \(s_ n\to s\), (P), to mean that \(\sum^ \infty_{k=0}s_ kp_ kt^ k/p(t)\to s\) as \(t\to R\). It is no restriction to assume \(R=1\) or \(R=\infty\). The method P is regular if and only if \(p(1)=\infty\) when \(R=1\) and \(p(t)\) is not a polynomial when \(R=\infty\). The author assumes this regularity together with conditions on orderly behaviour of the sequence \(p_ n\) and then proves the following `almost sure' summability theorem. Let \(X_ 0\), \(X_ 1,\dots\) be i.i.d. random variables. Then (i), (ii) and (iii) are equivalent: (i) a moment condition on \(X_ 0\); (ii) \(X_ n-C_ n\to 0\), (P), almost sure, where \(C_ n=EX_ 0\) when \(E| X_ 0|<\infty\) and \(C_ n\) is a truncated moment of \(X_ 0\) when \(E| X_ 0|=\infty\); (iii) \(X_ n-C_ n\to 0\), (S), where S is a summability method implying P. Depending on the behaviour of \(p_ n\), the method S is convergence of the means \(\sum^ n_{k=0}p_ ks_ k/\sum^ n_{k=0}p_ k\) or of certain moving averages of the \(s_ k\). As corollaries almost sure summability theorems follow for generalized logarithmic, Abel, Borel and \(\text{B}_ p\) methods, with S a Riesz typical means method that may reduce to Cesàro \(\text{C}_ 1\).











This page was built for publication: Power series methods and almost sure convergence

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5288373)