No label defined (Q3680083)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3901825
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 3901825

      Statements

      1985
      0 references
      non-identically distributed random variables
      0 references
      inequalities
      0 references
      moments
      0 references
      record times
      0 references
      asymptotically independent uniform variables
      0 references
      Markov chain
      0 references
      Markov property
      0 references
      inter-record times
      0 references
      0 references
      For a sequence of random variables \(X_ 1,X_ 2,..\). the record times L(n), \(n\geq 1\), are defined as \(L(1)=1\), and, for \(n\geq 2\), \(L(n)=\min \{j: j>L(n-1)\), \(X_ j\geq X_{L(n-1)}\}\). When the \(X_ j\) are independent and identically distributed, the asymptotic theory of L(n) is well developed [see chapter 6 in the reviewer's book ''The asymptotic theory of extreme order statistics.'' (1978; Zbl 0381.62039)]. \textit{M. C. K. Yang} [J. Appl. Probab. 12, 148-154 (1975; Zbl 0306.60068)] extended several results on L(n) to the case when the \(X_ j\) are independent, and the distribution function of \(X_ j\) is \(F^{a(j)}(x)\), where F(x) is a continuous distribution function, and a(j) is a sequence of numbers increasing at the rate of \(a^ j\) with some number \(a>1\). The author generalizes this latter case by permitting the sequence a(j) to be generated by a continuous function S(x) by the rule \(S(n)=a(1)+a(2)+...+a(n).\) Several remarkable results are proved. For example, E[L(2)] is finite if, and only if, \(\int^{+\infty}_{1}[1/S(x)]dx<+\infty.\) Furthermore, if \(S(x)\to +\infty\) and \(S(x+1)/S(x)\to 1\) as \(x\to +\infty,\) then the random variables \(S[L(n+t+1)]/S[L(n+t)]\) are asymptotically independent uniform variables on (0,1), where \(n\to +\infty\) and \(1\leq t\leq k\) with k fixed. It is shown that L(n), \(n\geq 1\), form a Markov chain, and just as in the identically distributed case [the reviewer and \textit{E. Seneta}, Proc. Am. Math. Soc. 50, 383-387 (1975; Zbl 0276.60028)], the Markov property is utilized with success for the investigation of the so called inter-record times \(\Delta (n)=L(n)-L(n-1).\) Note that, by the choice \(S(x)=x\), the results known for the identically distributed case are also reobtained.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references