On Jiang's asymptotic distribution of the largest entry of a sample correlation matrix (Q444980): Difference between revisions

Let \(\{ X,X_{k,i};\;i \geq 1,\;k \geq 1\}\) be a double array of nondegenerate independent identically distributed random variables. Let \(\{p_n,\;n \geq 1\}\) be a sequence of positive integers such that \(n/p_n\) is bounded away from zero. Let \(\hat \rho_{i,j}^{(n)}\) denote the correlation coefficient between \((X_{1,i},\dots,X_{n,i})^\prime\) and \((X_{1,j},\dots,X_{n,j})^\prime\) and \(L_n=\max|\hat \rho_{i,j}^{(n)}|.\) The authors solve a problem posed by\textit{D. Li, W.-D. Liu} and \textit{A. Rosalsky} [Probab. Theory Relat. Fields 148, No. 1--2, 5--35 (2010; Zbl 1210.62010)]. Under the assumption \(EX^2<\infty,\) they show that the following three statements are equivalent: \[ \lim_{n \rightarrow \infty} n^2 \int_{(n \log n)^{1/4}}^\infty (F^{n-1}(x)-F^{n-1}(\frac{\sqrt{n\log n}}{x}))dF(x)=0;\tag{1} \] \[ (\frac{n}{\log n})^{1/2}L_n \text{ converges in probability to 2 as } n \rightarrow \infty;\tag{2} \] and \[ \lim_{n \rightarrow \infty}P(nL_n^2-a_n\leq t)=\exp[-\frac{1}{\sqrt{8\pi}}e^{-t/2}],\quad -\infty <t<\infty;\tag{3} \] where \(F(x)= P(|X|\leq x),\;x \geq 0\) and \(a_n= 4 \log p_n-\log\log p_n,\;n \geq 2.\)