Estimates of the rate of convergence for max-stable processes (Q1124201)

Let \(\bar X=\{X_ j\); \(j\in {\mathbb{N}}\}\), \(\bar Y=\{Y_ j\); \(j\in {\mathbb{N}}\}\) be the sequences of r.v.'s taking values in spaces \(L_ r(T)\), \(1\leq r<\infty\), C(T) or \(R_+^{\infty}\) and \({\mathcal C}\) be the class of all sequences \(\bar C=\{c_ j(n)\); \(j,n\in {\mathbb{N}}\}\) satisfying the conditions: \[ c_ 1(n)>0,\quad c_ j(n)\geq 0,\quad j=2,3,...,\quad \sum^{\infty}_{j=1}c_ j(n)=1. \] For any \(\bar X,\) \(\bar Y\) and \(\bar C\) define the normalized maxima \[ \bar X_ n:=\max_{j\in {\mathbb{N}}}c_ j(n)X_ j \] and \(\bar Y_ n\) analogously. Estimates of the deviation \(\mu(\bar X_ n,\bar Y_ n)\) with respect to a given metric \(\mu\) are considered. In particular, if the sequence \(\bar X\) consists of i.i.d. r.v.'s estimates in terms of the minimal metric \({\hat \mu}\) defined by \[ {\hat\mu}(X,Y)=\inf \{\mu(X',Y'):\quad X'=^{d}X,\quad Y'=^{d}Y\} \] are derived. Specifically, assuming that \(Y_ n=^{d}Y_ 1\) \(\forall \bar C\in {\mathcal C}\), i.e. \(Y_ 1\) is simple max-stable, and \[ a_ p(n)=[\sum^{\infty}_{j=1}c^ p_ j(n)]^{\bar p}\quad for\quad p\in (0,\infty),\quad \bar p= \begin{cases} 1 &\text{ for }p\leq 1,\\ 1/p &\text{ for }p>1, \end{cases} \] for some metrics \({\hat \mu}\) the estimates \[ {\hat \mu}(X_ n,Y_ n)\leq a_ p(n){\hat \mu}(X_ 1,Y_ 1) \] are proved. In contrast to the summation scheme, uniform estimates are also derived.