2007/02/06 by Joachim Krug · 1 citation
Decision Sciences · Mathematics · Physics and Astronomy · #Probability and Risk Models #Statistical Distribution Estimation and Applications #Stochastic processes and statistical mechanics #cond-mat.dis-nn #cond-mat.stat-mech #math.PR
paper · pdf · doi:10.1088/1742-5468/2007/07/p07001
published as J. Stat. Mech. (2007) P07001 · 12 pages, 2 figures
arxiv created 2007/02/06 · openalex publication_date 2007/07/03 · arxiv updated 2009/12/01 · openalex created_date 2016/06/24 · openalex updated_date 2026/08/01
In the context of this paper, a record is an entry in a sequence of random variables (RVs) that is larger or smaller than all previous entries. After a brief review of the classic theory of records, which is largely restricted to sequences of independent and identically distributed (i.i.d.) RVs, new results for sequences of independent RVs with distributions that broaden or sharpen with time are presented. In particular, we show that when the width of the distribution grows as a power law in time n , the mean number of records is asymptotically of order ln n for distributions with a power law tail (the Fréchet class of extreme value statistics), of order (ln n ) 2 for distributions of exponential type ( Gumbel class ), and of order n 1/(ν+1) for distributions of bounded support ( Weibull class ), where the exponent ν describes the behaviour of the distribution at the upper (or lower) boundary. Simulations are presented which indicate that, in contrast to the i.i.d. case, the sequence of record breaking events is correlated in such a way that the variance of the number of records is asymptotically smaller than the mean.