Renorming divergent perpetuities (Q638760): Difference between revisions

Let \((M_k, Q_k)_{k\geq 1}\) be independent copies of a random vector \((M,Q)\) with \(\operatorname{E}\log|M|\geq 0\). Under the latter condition and a couple of technical assumptions, the sequence \[ |X_n|:=\bigg|\sum_{k=1}^nQ_k\prod_{j=1}^{k-1}M_j\bigg| \] diverges to \(\infty\) in probability. The authors exhibit several cases in which \(X_n\), properly normalized and centered, weakly converges to a proper and nondegenerate probability law. A typical additional assumption in the present paper is that either \(|\log M|\) or \(Q\), or both have finite variance. It would be interesting to investigate the problem under less stringent assumptions. Reviewer's remark: The list of references does not contain the paper \textit{A. K. Grincevicius} [Lith. Math. J. 15(1975), 568--579 (1975; Zbl 0372.60045)] which is one of the most important works concerning divergent perpetuities.