Domains of attraction and regular variation in \({\mathbb{R}}^ d\) (Q791974): Difference between revisions

The authors define an extended form of regular variation of functions on \({\mathbb{R}}^ 2_+\). The function f: \({\mathbb{R}}^ 2_+\to {\mathbb{R}}_+\) varies regularly at infinity with auxiliary functions r: \({\mathbb{R}}_+\to {\mathbb{R}}_+\) and s: \({\mathbb{R}}_+\to {\mathbb{R}}_+\) if \(f(xr(t),ys(t))/f(r(t),s(t))\to \lambda(x,y), x>0\), \(y>0\), as \(t\to \infty\). For f monotone in each variable and r,s both varying regularly at \(\infty\), they prove the extensions to \({\mathbb{R}}^ 2_+\) of the well known theorems for regular variation in \({\mathbb{R}}_+:\) Karamata's integral theorem, regular variation of a monotone derivative of f and the Abel-Tauber theorem connecting the behaviour of f at infinity and of the Laplace transform of f at zero. These results are applied to the stable attraction of random vectors in \({\mathbb{R}}^ 2:\) Let \((X_ i,Y_ i)\) be i.i.d. random vectors. If \((a_ n^{-1}\sum^{n}_{k=1}X_ k-c_ n\), \(b_ n^{- 1}\sum^{n}_{k=1}Y_ k-d_ n)\) converges in distribution to nondegenerate (V,W), the distribution of (V,W) has to be stable. The possible stable limit distributions and conditions for convergence were determined by \textit{S. Resnick} and \textit{P. Greenwood}, ibid. 9, 206-221 (1979; Zbl 0409.62038). For \(X_ i>0\), \(Y_ i>0\) the authors give new proofs of some of these results and also new conditions for convergence in terms of regular variation of \(P(X_ 1>x\), \(Y_ 1>y)\) and its integral. Finally, they succeed in transferring the convergence theorems to not necessarily positive \(X_ i\) and \(Y_ i\), if (V,W) or V is normal.