A theorem of Hsu-Robbins type for negatively associated sequence (Q1919008)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | A theorem of Hsu-Robbins type for negatively associated sequence |
scientific article |
Statements
A theorem of Hsu-Robbins type for negatively associated sequence (English)
0 references
27 January 1997
0 references
Random variables \(X_1, \dots, X_n\) are said to be NA (negatively associated) if for every pair of disjoint nonempty subsets \(T_1\), \(T_2\) of \(\{1, \dots, n\}\), \[ \text{Cov} \bigl (f_1(X_i,i \in T_1),\;f_2 (X_j,j \in T_2) \bigr) \leq 0, \] where \(f_1\) and \(f_2\) are coordinate-wise nondecreasing functions. A sequence of r.v.'s \(\{X_i, i \geq 1\}\) is said to be NA if for every \(n \geq 2\), \(X_1, \dots, X_n\) are NA. A theorem of Hsu-Robbins type for the NA sequence is established in this paper: Suppose that the sequence of r.v.'s \(\{X_i, i \geq 1\}\) is NA with the same univariate marginal distribution, and let \[ S_n = \sum^n_{i = 1} X_i, \qquad S_n^{(k)} = \sum^n_{i = 1,i \neq k} X_i = S_n - X_k, \quad k = 1, \dots, n. \] Then \[ \sum^\infty_{n = 1} P \Bigl( \max_{1 \leq k \leq n} |S_n^{(k)} - nb |\geq n \varepsilon \Bigr) < \infty \] for every \(\varepsilon > 0\) iff \(EX_1 = b\) and \(EX^2_1 < \infty\). We also obtain the following corollary: Suppose that the sequence of r.v.'s \(\{X_i, i \in N\}\) is NA with the same univariate marginal distribution, and let \(S_n = \sum^n_{i = 1} X_i\). If \(EX^2_1 < \infty\) and \(EX_1 = b\), then \[ \sum^\infty_{n = 1} P \bigl( |S_n - nb |\geq n \varepsilon \bigr) < \infty \text{ for every } \varepsilon > 0. \] We establish the above results in a new way due to the NA specific properties. Two interesting examples are also given to show that we must pay attention to the properties of mixed moments of high-order for the NA r.v.'s.
0 references
complete convergence
0 references
properties of mixed moments of high-order
0 references
negatively associated
0 references