On the influence of weights on extremes (Q5926466)

\textit{P. Deheuvels, D. M. Mason} and \textit{G. R. Shorack} [Ann. Inst. Henri Poincaré, Probab. Stat. 29, No. 1, 83--103 (1993; Zbl 0774.62042)] proved the following theorem: Assume that \(m(n)\to+\infty\) and \((m(n)/n)\to 0,\) as \(n\to+\infty.\) Then, as \(n\to+\infty\), \[ {\mathbf P} (m(n)([U_{n,n}-U^{*}_{m(n),m(n)}]/U_{n,n})>x |U_{1},\dots,U_{n})\to\exp(-x),\;x\geq 0, \] holds in probability. In addition, if \[ ((m(n)\log\log n)/n)\to 0, \] as \(n\to+\infty,\) then it holds with probability 1. Here: \((U_{k};k\geq 1)\) are sample independent uniform \((0,\theta),\) \(\theta> 0,\) r.v.s, where \(\theta\) is an unknown parameter; \((U_{k,n}; k=1,\dots,n)\) are order statistics of these variables; \((U^{*}_{k}; k=1,\dots,n)\) is a sample from the empirical distribution function \({\mathbf F}_{n}\) of \(U_{n};\) \((U^{*}_{k,n}; k=1,\dots,n)\) are order statistics of the \(U^{*}_{n}.\) The aim of this paper is to establish similar results when weights are introduced. The motivation of this paper is that the ``weights'' choice corresponds to a practical situation. For example, ratios are commonly used in poll or insurance to associate weights to the data. The influence of the maxima in the construction of limit theorems is studied under suitable assumptions on the weights.