Two theorems on the sequence of maxima of independent random variables (Q1068433)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Two theorems on the sequence of maxima of independent random variables
scientific article

    Statements

    Two theorems on the sequence of maxima of independent random variables (English)
    0 references
    0 references
    0 references
    1984
    0 references
    Let: a) \(\{\xi_{n\kappa}\}\) be a sequence of i.i.d. random variables with a d.f. \(F_ n(x)\), for fixed n, and every \(\kappa\) ; b) \(\{\kappa_ n\}\) be a sequence of positive integers; c) \(\{\nu_ n\}\) be a sequence of non-negative integer random variables, independent of \(\{\xi_{n\kappa}\}\). Let \(\eta_{\kappa_ n}=\max\) \((\xi_{n_ 1},...,\xi_{n\kappa_ n})\) and \(\eta_{\nu_ n}=\max (\xi_{n1},...,\xi_{n\nu_ n}).\) It was proved previously by \textit{B. V. Gnedenko} and \textit{D. B. Gnedenko} [Serdica 8, 229-234 (1982; Zbl 0507.60024)] that if A) \(P(\eta_{\kappa_ n}<x)\to \Phi (x)\) and B) \(P(\nu_ n/\kappa_ n<x)\to A(x)\) (where \(\Phi\) (x) and A(x) are d.f.'s) then C) \(P(\eta_{\nu_ n}<x)\to \psi (x)=\int^{+\infty}_{0}(\Phi (x))^ zdA(z).\) The authors are concerned with possible converses of this theorem. They prove, first, that if B) and C) are valid and A(x) is not the d.f. of the almost sure zero random variable then A) is also valid. Furthermore, a weaker form of the other converse is shown: if A) is substituded by the special case A'), i.e. if we have a sequence of i.i.d. random variables \(\{\xi_ n\}\) with d.f. F(x) and \(\sigma_ n=\max (\xi_ 1,...,\xi_ n)\) and we define \(\xi_{n\kappa}=(\sigma_{\kappa}-b_ n)/a_ n\), take \(\kappa_ n=n\), and \(P((\sigma_ n-b_ n)/a_ n)<x)\to \Phi (x)\) (which can be only Weibull, Gumbel or Frechet d.f.'s for maxima), then A') and C) imply B).
    0 references
    limit distributions of maxima
    0 references
    maxima of samples with random size
    0 references

    Identifiers