No label defined (Q3983595)

By using the method proposed by the author in his earlier works, sum distribution problems for sequences and triangular arrays with mixing are studied, as well as maximum distribution. The approach is based on the partition of a sum into three parts, on the weak influence of the second term on the whole sum, weak dependence of the first and third parts and suitable choice of approximating families of independent random variables for these parts. After the general description of the method the case of convergence to an infinite divisible distribution is discussed when \(\rho\)-mixing holds. Then the estimates for the second moment of the norm of a sum of random elements and central limit theorem in the space of type 2 in the case of \(\varphi\)-mixing are given. The estimate of the rate of convergence is constructed for the moments in the central limit theorem for Banach spaces in the case of \(\varphi\)-mixing. Finally, an array \(\{X_{1,n},\ldots, X_{n,n},n\geq 1\}\) of identically distributed for each fixed \(n\) random variables is considered and the following limit theorem for \({\mathcal M}_{1,n}\) is proved where \({\mathcal M}_{r,s}=\max\{X_{i,n}:r\leq i\leq s\}\): If \(\sum^ \infty_{k=1}q(2^ k)<\infty\) with \[ q(k)=\sup\{| P({\mathcal M}_{s+k,n}\geq t\mid{\mathcal M}_{1,s}\geq t)-P({\mathcal M}_{s+k,n}\geq t)|:s\geq 1,\;n\geq 1,\;t\in R^1\}, \] for any \(\tau\in R^1\) there exists \[ \lim_n nP(X_{1,n}\geq\tau)=\beta(\tau)\quad\text{and }\max\{P(X_{a+j,n}\geq\tau\mid X_{a,n}\geq\tau):\;1\leq a\leq n-j\}\to 0 \] as \(n\to\infty\) for any natural \(j\) and for \(\tau\) such that \(0<\beta(\tau)<\infty\), then \[ P({\mathcal M}_{1,n}<\tau)\to e^{- \beta(\tau)} \] as \(n\to\infty\) for any \(\tau\) for which \(\beta(\tau)<\infty\). Compare with \textit{R. M. Loynes} [Ann. Math. Stat. 36, 993--999 (1965; Zbl 0178.53201)].