Limit theorems for the maximum terms of a sequence of random variables with marginal geometric distributions (Q5943417): Difference between revisions

The author considers the following problems. Let \(X_{n1}^\ast, \ldots, X_{nn}^\ast\) be a sequence of \(n\) independent random variables which have a geometric distribution with the parameter \(p_n ={1\over n}\) and let \(M_n^\ast = \max\{ X_{n1}^\ast, \ldots, X_{nn}^\ast\}.\) Let \(Z_1, Z_2, \ldots\) be a sequence of independent random variables with uniform distributions over the set \(N_n =\{ 1, 2, \ldots, n\}.\) Let \(X_{nj} =\min \{ k\colon Z_k =j\}\), \(M_n = \max \{ X_{n1}, \ldots, X_{nn}\}.\) The limiting distributions of \(M_n\) and \(M_n^\ast\) are found. The author proves that \[ \lim_{n\to\infty} P \{ M_n \leq n(x + \ln n)\} =\exp\{ -e^{-x}\}\quad\text{and}\quad\lim_{n\to\infty} P \{ M_n^\ast \leq n(x + \ln n)\} =\exp\{ -e^{-x}\}. \] The rates of convergence of \(M_n\) and \(M_n^\ast\) are found. Let \(S_n\) be \(2\)nd largest among \(X_{n1}, \ldots, X_{nn}.\) The limiting bivariate distribution of \((S_n, M_n)\) is obtained. Let \((\alpha_n)\) and \((\beta_n)\) be two sequences of positive integers such that \(\alpha_n, \beta_n \in N_n\), \[ \begin{gathered} A_n =\{ 1,2,\dots, \alpha_n \},\quad B_n = \{ n-\beta_n +1, n-\beta_n +2, \dots, n\},\\ M(A_n) = \max \{ X_{n1}, X_{n2},\dots, X_{n \alpha_n}\}, \\ M(B_n) = \max \{ X_{n,n-\beta_n +1}, X_{n,n-\beta_n +2}, \dots, X_{n n}\},\quad T_n = \min \{ M(A_n), M(B_n) \}.\end{gathered} \] The limiting distributions of \(M(A_n)\) and \(T_n\) are found.