complete convergence for arrays of minimal order statistics数组成套收敛最小次序统计量.pdfVIP

IJMMS 2004:44, 2325–2329 PII. S0161171204401379 © Hindawi Publishing Corp. COMPLETE CONVERGENCE FOR ARRAYS OF MINIMAL ORDER STATISTICS ANDRÉ ADLER Received 26 January 2004 For arrays of independent Pareto random variables, this paper establishes complete conver- gence for weighted partial sums for the smaller order statistics within each row. This result improves on past strong laws. Moreover, it shows that we can obtain a finite nonzero limit for our normalized partial sums under complete convergence even though the first moment of our order statistics is infinite. 2000 Mathematics Subject Classification: 60F15. Let {Xnj , 1 ≤ j ≤ mn , n ≥ 1} be independently distributed random variables with density fXnj (x) = pnx −pn −1I (x ≥ 1), where pn 0. Let Xn(k) be the kth smallest order statistic from each row of our array. Thus the density of Xn(k) is pn ·mn ! −p (m −k+1)−1 −p k−1 n n n fXn(k) (x) = (k − 1)! mn −k !x 1 −x I (x ≥ 1). (1) We will establish laws of large numbers of the form ∞ N a X n=k n n(k) cN P −L ∞ (2) bN N =k for all 0, where L is not zero even though EXn(k) = ∞ and of cours