Rarely Observed Sample Maxima
The class MSS of the max-semistable df's on \( R^1 \) can be characterized by the following equivalent statements. For \( \alpha \in (0,1) \) and \( L(x) = ax+b \) let \( G \in \text{MSS}(\alpha, L) \). Then i) \( G \) satisfies the functional equation \[ G^{\alpha}(x) = G(L(x)), \] ii) there exist df \( F \) and norming sequences \( \{ a_n \}\), \(a_n>0\), \(\{ b_n \}\), \(b_n \) real, such that \[ G(x) = \lim F^{k_n}( a_nx + b_n), \quad k_n/k_{n+1} \rightarrow \alpha,\;n \rightarrow \infty, \] iii) there is a positive bounded periodic function \( p(x) \) with a period \( T = \log \frac{1}{\alpha} \) such that \[ G(x) = \exp \{ -e^{h(x)}p(h(x))\}. \] The function \( h(x) \) is in fact \(-\log (-\log H(x)) \) where \( H \) is a max-stable df. The properties of the univariate MSS-df's and their domain of attraction have been studied by \textit{I. Grinevich} (1993) in her Ph.D. thesis. The multivariate MSS-df's under nonlinear normalisations and their integral representation have been investigated by \textit{E. Pancheva} [Theory Probab. Appl. 38, No. 3, 551--556 (1993) and Teor. Veroyatn. Primen. 38, No. 3, 679--684 (1993; Zbl 0816.60024)]. The most recent results in this field are due to \textit{Z. Megyesi} (2002) who gave also a.s. limit theorems. In the present paper the authors consider a subclass of the max-semistable df's for which \( \alpha = e^{-d}\), \(d>0 \), namely the ``rarely observed sample maxima. Let \(s_n = 1+a+a^2+\dots +a^{n-1}\), \(L(x)=ax+1 \) and let \( L^{\circ n} \) be the \(n\)th convolution of \( L \). Then \( L^{\circ n}(x) = a^nx + s_n \) and an \(n\)-times iteration of i) gives the functional equation \[ G^{\alpha^n}(x) = G(L^{\circ n}(x)), \quad n=\dots,-2,-1,0,1,2,\dots \] In Theorem 1 the authors show that \( G \in \text{MSS}(e^{-d}, L) \) iff for \(x\in[0,1]\) \[ -\log(-\log G(a^nx + s_n)) = nd + y(x) \] where the function \(y:[0,1] \rightarrow [0,d] \) is non-decreasing, right continuous and \( y(x)>0 \) for \(x>0\). In Theorem 2 the domain of attraction of \( G \) is discussed in terms of the functions \( v(x):= -\log(-\log F)^{\leftarrow}(x) \) and \( \psi (x) = -\log(-\log G)^{\leftarrow}(x) \).
- Statistics of extremes for IID data and breakthroughs in the estimation of the extreme value index: Laurens de Haan leading contributions
- On the max-semistable limit of maxima of stationary sequences with missing values
- Limit laws for maxima of a stationary random sequence with random sample size
- Existence of multivariate max-universal laws
- Maxima of random properties of particles in supercritical branching processes
- On Max-Multiscaling Distributions as Extended Max-Semistable Ones
- On the maximum of periodic integer-valued sequences with exponential type tails via max-semistable laws
- On Stochastic Evolution Equations with Stochastic Boundary Conditions
- Extreme Value Theory and Statistics of Univariate Extremes: A Review
- Max-semistable extreme value laws for autoregressive processes with Cantor-like marginals
- Asymptotic distribution of certain statistics relevant to the fitting of max-semistable models
- Generalized Pickands' estimators for the tail index parameter and max-semistability
This page was built for publication: Rarely Observed Sample Maxima
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4328510)
