ERDS-RE'NYI大数律下自正则的效用

ON THE EFFECT OF SELF NORMALIZATION WITH ERDS RE'NYILAW OF LARGE NUMBERS

｛Ｘ，Ｘｉ，ｉ≥１｝是ｉ．ｉ．ｄ．ｒ．ｖ′．ｓ．在矩母函数存在的条件下，由古典的Ｅｒｄｏｓ－Ｒéｎｙｉ大数律有ｌｉｍｎ→∞ｍａｘ０≤ｋ≤ｎ∑ｋ＋［ｃｌｏｇｎ］ｉ＝ｋ＋１Ｘｉ［ｃｌｏｇｎ］＝α（ｃ），α（ｃ）为某常数．自正则下ＭｉｋｌóｓＣｓｏｒｇｏ＆ＳｈａｏＱｉｍａｎ（１９９４）在仅要求一阶矩的条件下就得到了：ｌｉｍｎ→∞ｍａｘ０≤ｋ≤ｎ∑ｋ＋［ｃｌｏｇｎ］ｉ＝ｋ＋１Ｘｉ∑ｋ＋［ｃｌｏｇｎ］ｉ＝ｋ＋１（Ｘ２ｉ＋１）＝β（ｃ），β（ｃ）为某常数．众所周知，自正则下人们往往在较弱条件下取得相应结果是因为：分母中的Ｘ能有效抵销分子中Ｘ较大而引起整个分式极限行为的波动．因此，在什么样的条件下，式ｍａｘ０≤ｋ≤ｎ∑ｋ＋［ｃｌｏｇｎ］ｉ＝ｋ＋１Ｘｉ∑ｋ＋［ｃｌｏｇｎ］ｉ＝ｋ＋１Ｘ２ｉ１－β［ｃｌｏｇｎ］β→ｒ（ｃ）成为非常有意思的问题，因为它将依赖于β的大小．本文给出，当０＜β≤１２时，只要Ｅ（Ｘ）≥０，上式就有有限极限．当１２＜β＜１时，则必须在矩母函数存在下，上式才有有限极限．并都求出了其极限表达式．

X,X i,i≥1} is i.i.d.r.v′.s, It is well known that  lim n→∞ max 0≤k≤n∑k+[c log n]i=k+1X i[c log n]=α(c),(α(c) is a constant depending on c), when E(e t 0X )<∞, t 0>0. By normalizing, paper got  lim n→∞ max 0≤k≤n∑k+[c log n]i=k+1X i∑k+[c log n]i=k+1(X 2 i+1)=β(c),(β(c) is a constant) only under the condition of one moment. In this paper it is studied that under what conditions one can gets  lim n→∞ max 0≤k≤n∑k+[c log n]i=k+1X i∑k+[c log n]i=k+1X 2 i 1-β [c log n] β→r(c), (r(c) is a constant) when β varies from 0 to 1, the problem has been solved completely.